Community and International Nutrition

Community and International Nutrition Intakes of Calcium and Vitamin D Predict Body Mass Index in the Population of Northern Norway 1 Elena Kamycheva, 2 Ragnar M. Joakimsen and Rolf Jorde Medical Department, University Hospital of North Norway, Tromsø, Norway ABSTRACT This study was designed to investigate whether there is any association between body mass index (BMI) and life-style factors, with a special emphasis on calcium and vitamin D intakes. In 1994/1995, men (n 9252) and women (n 9662) participated in the fourth Tromsø study and completed the food-frequency and life-style factor questionnaires. We measured BMI, coffee and alcohol consumption, physical activity, smoking, and calcium and vitamin D intakes. A negative association between physical activity, smoking and BMI, and a positive association between BMI and coffee intake were found in both sexes (P 0.001). BMI and calcium intake were positively related in men (P 0.001), but not in women. BMI and vitamin D intake were negatively associated in both sexes (P 0.001), which to our knowledge has not been reported before. The lowest quartile of vitamin D intake was an independent predictor of obesity (defined as BMI 30 kg/m 2 ) in men and women (P 0.001 in both genders), resulting in odds ratios of 2.24 [95% confidence interval (CI) 1.80, 2.80] in men and 1.51 (95% CI 1.23, 1.85) in women compared with the highest quartile. In conclusion, calcium and vitamin D intakes may have opposing effects on body weight, which is difficult to explain given current knowledge of calcium metabolism. J. Nutr. 132: 102 106, 2002. KEY WORDS: vitamin D obesity calcium life-style factors Obesity is a predisposing factor for poor health and allcause mortality (1), and an increase in body weight has occurred worldwide (2). In our region, Northern Norway, there has been an increase in mean body mass index (BMI) 3 from 1974 to 1994/1995 of 1 kg/m 2 in both sexes (3). The cause of obesity is both genetic and environmental (4). However, the increase that has taken place recently is the result of life-style changes with a decrease in physical activity and an increase in energy intake observed in most societies (2). Not only the amount, but also the type of food ingested is important. Thus, there have recently been a few reports that a high intake of calcium is associated with a low body weight (5,6), which, if causal, may affect the composition of weight-reducing diets. There have been several large health surveys in Tromsø; the last one in 1994 1995 included 27,000 subjects. Questions on life-style factors, as well as a food-frequency questionnaire that included questions on calcium and vitamin D intake were part of the study, thus providing a large database for testing the hypothesis on a relation between intake of calcium and vitamin D and overweight. 1 Supported by a grant from the research program Research on the Elderly in Tromsø, which is funded by the Norwegian Ministry of Health and Social Affairs. 2 To whom correspondence should be addressed. E-mail: elena.kamycheva@unn.no. 3 Abbreviations used: BMI, body mass index; CI, confidence interval; PTH, parathyroid hormone; RDA, recommended dietary allowance; -GT, -glutamyl transferase. SUBJECTS AND METHODS The present cross-sectional study, which took place in 1994 1995, was part of the fourth Tromsø study and was conducted similarly to the previous studies (7). Men and women 24 y old were given a questionnaire concerning medical history, smoking habits, coffee and alcohol intake, and physical activity. The number of cigarettes smoked daily was recorded. An alcohol intake score was calculated by totaling the number of glasses of beer, wine, and spirits per 14 d (assuming an equal amount of alcohol in one glass of each type). A physical activity score was calculated by totaling the hours of moderate and hard physical activity (giving the hours with hard activity double the value). The food-frequency questionnaire included questions on intake of milk, bread, butter on bread, yogurt, cheese and eggs, how often they consumed fatty fish for dinner (e.g., salmon), and use of fatty fish (e.g., mackerel or sardines) in sandwiches. Furthermore, the subjects were asked about their daily use of calcium and vitamin D tablets and cod liver oil supplementation. The food-frequency questionnaire was not designed to allow calculation of energy intake and dietary fat. Height and weight were measured while the subjects wore light clothing and no shoes. Blood samples were drawn from nonfasting subjects and serum -glutamyl transferase ( -GT) was measured with a spectrophotometric method using the multianalyzer Hitachi 917 (Roche Diagnostics, Basel Switzerland) with reagents from Boehringer Mannheim (Mannheim, Germany). The Regional Ethical Committee approved the study, and all participants gave informed consent. To check the linearity, we recoded age, vitamin D intake, calcium intake and physical activity into quartiles. In the linear regression model, with BMI as the dependent variable, these covariates were each entered as three dummy variables with the first group as the reference category for the initially continuous variables. Age and physical activity sustained linearity to BMI in both genders, whereas vitamin D intake had linear relation to BMI only in women and calcium intake only in men. Therefore, vitamin D intake in men and calcium intake in women were entered into the final linear model as dummy variables (8). We analyzed men and women separately with linear regression using BMI as the dependent variable, and physical activity, alcohol 0022-3166/02 $3.00 2002 American Society for Nutritional Sciences. Manuscript received 2 May 2002. Initial review completed 28 May 2002. Revision accepted 26 September 2002. 102

VITAMIN D INTAKE AND BODY MASS INDEX 103 and coffee consumption, smoking, and calcium and vitamin D intake as independent variables. The dependent variable (BMI) appeared to be normally distributed on visual inspection and also had a skewness of 1.0 0.018. Differences between groups were evaluated with one-way ANOVA and independent samples t test. Where appropriate, Bonferroni correction was applied. Linear correlations were evaluated with the Pearson correlation coefficient. Vitamin D intake in both genders was divided into 4 quartiles ( 2.8, 2.8 5.8, 5.9 10.0, 10.0 g/d), and calcium intake in women into 4 quartiles ( 275, 275 381, 382 595, 595 mg/d), to create dummy variables and enter them into the linear model with the lowest quartile as the reference category. Obesity was defined as a BMI 30 kg/m 2. We applied a logistic regression model with obesity as the dependent variable and vitamin D intake quartiles as predictors (entered as dummy variables, thus not assuming linearity or exponentiality), adjusted for age, calcium intake, physical activity, coffee and alcohol consumption, and smoking. From this model, the odds ratio was calculated for those in the lowest vitamin D intake quartile compared with those in the highest vitamin D quartile as a reference category. To evaluate the potential public health role of vitamin D, we calculated the predicted BMI in our population if everyone consumed the recommended daily allowance (RDA) dose of vitamin D (5 g/d for those 50 y old and 10 g/d for those 50 y) (9). The adjustment of BMI for those who consumed less than the RDA dose of vitamin D was made according to the following formula: BMI P BMI O b (vitamin D RDA vitamin D O ) For those who already consumed the RDA dose of vitamin D or higher: BMI P BMI O where BMI P is the predicted BMI, BMI O is the observed BMI from the Tromsø study, b is the unstandardized regression coefficient from the linear model (vitamin D intake coded as a continuous variable in women, coded as the lowest quartile or not in men), vitamin D RDA is the recommended daily allowance dose of vitamin D and vitamin D O is the observed intake from the Tromsø study. Differences with P 0.05 were considered significant. All tests were two-sided. The data were analyzed with SPSS for Windows, version 10.0 (SPSS, Chicago, IL). RESULTS A total of 12,866 men and 14,293 women were examined, which constituted 74.2 and 79.0%, respectively, of the eligible population. Because only subjects aged 70 y answered the complete diet questionnaire, this study was restricted to those below this age. Furthermore, not all subjects answered all of the relevant questions, thus leaving 9252 men and 9662 women to be included in the following analysis; they comprised 79.5 and 77.0%, respectively, of those 70 y old who participated in the Tromsø study. There were no differences in age, BMI or gender distribution between these subjects and those not included in the final analyses (data not shown). The demographics of the study population are given in Table 1; the table also gives the correlation coefficients and standardized regression coefficients for BMI vs. the other variables. In the multiple linear regression model, age was the most significant predictor of BMI (P 0.001). In both sexes, BMI and coffee intake were positively associated (P 0.05) and in women, BMI and alcohol intake were negatively associated (P 0.001). The alcohol intake score and -GT levels were positively correlated (r 0.143; P 0.001, men and women together). Calcium intake and BMI had no direct correlation, but after correcting for the other variables with age as the most important in the multiple regression analysis, there was a significant positive association in men (P 0.001). In women, no association was found even with the entry of categorical dummy variables. Vitamin D intake had a negative correlation with BMI in TABLE 1 Body mass index (BMI), age, coffee and alcohol consumption, physical activity, smoking, and calcium and vitamin D intakes in men and women, and their respective Pearson correlation coefficients (r) with BMI, and standardized regression coefficients ( ) from the linear model with BMI as dependent variable1 Men2 r Women2 r n 9252 9662 BMI, kg/m2 25.6 0.03 24.4 0.04 Age, y 43.9 0.12 0.15* 0.141* 42.8 0.12 0.21* 0.228* Coffee, ml/d 900.0 6.30 0.03** 0.028* 705.0 4.90 0.03** 0.043* Alcohol,3 units/14 d 6.3 0.07 0.02 0.007 3.0 0.04 0.07* 0.104* Physical activity, h/wk 4.4 0.03 0.08* 0.096* 3.8 0.02 0.09* 0.137* Smoking, cigarettes/d 2.9 0.05 0.08* 0.058* 2.5 0.04 0.07* 0.064* Calcium intake,4 mg/d 541.4 3.00 0.05* 0.016 437.0 2.45 0.019 Calcium intake, 0 274 mg/d ref.5 275 381 mg/d 0.02 382 595 mg/d 0.02 595 mg/d 0.00 Vitamin D intake,6 g/d 7.1 0.05 0.050* 6.5 0.05 0.05* 0.018 Vitamin D intake, 0 2.7 g/d ref.7 2.8 5.8 g/d 0.15* 5.9 10.0 g/d 0.13* 10.0 g/d 0.14* 1 R2 0.038 and 0.074 for men and women, respectively. 2 Values are means SEM. *P 0.001; ** P 0.05. 3 1 unit of alcohol 15 ml of spirits. 4 Calcium intake is entered here as a continuous variable for men. 5 Calcium intake is entered here as a dummy variable for women. 6 Vitamin D intake is entered here as a continuous variable for women. 7 Vitamin D intake is entered here as a dummy variable for men.

104 KAMYCHEVA ET AL. men, and after correction for age and other variables in the multiple regression analysis, a significant negative association was found in both sexes (P 0.001). Associations between vitamin D intake quartiles and BMI after adjusting for age were found for both genders (Fig. 1). Among those studied, 764 men and 826 women were obese. In the logistic regression model, the lowest quartile of vitamin D intake was a significant independent predictor of obesity (P 0.001) for both genders, resulting in odds ratios, adjusted for age, calcium intake, physical activity, coffee and alcohol consumption and smoking, of 2.24 [95% confidence interval (CI) 1.80, 2.80] in men and 1.51 (95% CI 1.23, 1.85) in women, with the highest quartile as the reference category. In addition, subjects who were obese consumed less vitamin D than nonobese subjects in all age groups (Figs. 2 and 3). The predicted BMI after ingestion of the RDA or higher dose of vitamin D was almost 0.20 kg/m 2 lower in men and 0.25 kg/m 2 lower in women in each age group. DISCUSSION Our main purpose was to examine the relationship between intakes of calcium and vitamin D and BMI. To our surprise we found a significant and positive association between calcium intake and BMI in men and negative and significant associations between intake of vitamin D and BMI in both sexes. Before discussing these findings in detail, we will review the other associations found between life-style factors and BMI. Physical activity is crucial for energy expenditure (10), and as expected, we found a negative correlation with BMI (11). In both sexes, smoking had a significant negative association with BMI, which is consistent with other studies (12,13). Only in women did we find an inverse relationship between alcohol intake and BMI. Other studies have shown a negative association between alcohol intake and BMI in both sexes, but in heavy drinkers only (12). Our questionnaire is likely not an accurate measure of alcohol intake. The correlation between alcohol intake and -GT activity was positive and significant, but the correlation coefficient was rather low, indicating doubtful validity of the self-reported data. FIGURE 2 Vitamin D intake in obese [body mass index (BMI) 30 kg/m 2 )] and nonobese (BMI 30 kg/m 2 ) men in different age groups; obese differs from nonobese (P 0.01) in all age groups. Values are means SEM, n 9252. We found a weak but significant positive relation with BMI for coffee intake in both sexes. Theoretically, coffee intake should reduce BMI because coffee consumption decreases the amount of fat tissue by elevating thermogenesis (14) and stimulating the sympathoadrenal system (15). However, previous reports on associations between coffee and BMI differ in their conclusions (16,17). Recently, Davies et al. (5) reported a negative association FIGURE 1 Age adjusted body mass index (BMI) in men and women in relation to quartiles of vitamin D intake. Values are mean SEM, n 9252 for men and n 9662 for women. P-values for difference between the upper and lower quartiles of vitamin D intake are 0.001 for men and 0.01 for women. FIGURE 3 Vitamin D intake in obese [body mass index (BMI) 30 kg/m 2 )] and nonobese (BMI 30 kg/m 2 ) women in different age groups; obese differs from nonobese (P 0.01) in all age groups. Values are means SEM, n 9662.

VITAMIN D INTAKE AND BODY MASS INDEX 105 between calcium intake and BMI in women. In their study, a meta-analysis of four separate studies on calcium supplementation for osteoporosis in which weight was also recorded, an increment in the calcium intake of 100 mg/d led to a 0.82 kg weight decrement (5). They also estimated that the calcium intake could explain 3% of body weight variability (5). Furthermore, in a trial in young women carried out by Teegarden et al. (18), subjects with a high calcium intake had less fat tissue and failed to gain weight, and in a study in young children, Skinner et al. (19) found that dietary calcium had a negative relation to body fat mass. As noted by Davies et al. (5), the lack of focus on calcium intake and weight can be explained by the lack of a plausible explanation for the association. However, there are some recent theories that link calcium metabolism to body weight. Thus, Zemel et al. (6) suggested that intracellular calcium enhances lipid deposition in adipocytes. The calciotrophic hormones, parathyroid hormone (PTH) and 1,25-dihydroxyvitamin D, promote calcium influx into the lipid cell, thus mediating lipogenesis. A low calcium intake increases these hormones and hence lipid storage as well (6,20). In support of this, there are repeated observations of a high BMI in subjects with elevated PTH (21,22). To our surprise, we could not confirm these observation about calcium intake and body weight. In our study, there was no association between BMI and calcium intake in women, whereas a positive and significant association was found in men. This is hard to explain. For both sexes, the calcium intake, which was calculated mainly from dairy sources, was remarkably low. This may indicate a substantial underreporting of calcium intake. The questionnaire was obviously less accurate than an interview with a dietician (23), although the reproducibility of dietary data from this self-administered questionnaire was evaluated previously and considered acceptable (24). In addition, our questionnaire cannot account for most of the variability in individuals who do not consume much calcium from dairy sources. Furthermore, a linear relationship between BMI and calcium intake did not exist in women, causing us to use dummy variables in the multiple regression model. These shortcomings could mask an association between calcium intake and BMI in women; in this respect, our findings must be viewed with great caution. On the other hand, this cannot explain the significant and positive association found in men. One possible explanation could be an association between calcium intake from dairy products and total energy intake, a factor for which we made no adjustments. Our food-frequency questionnaire was concerned mainly with calcium intake from dairy products, which in a Norwegian diet represent 75% of the total calcium intake (7); this is rather high compared with other populations (25). The relationship in men may therefore not be calcium specific, but could reflect a number of other components in dairy products. There might be a weight-increasing factor in dairy products that overrides a weight-reducing effect of calcium. Thus, in populations in which calcium intake is not based on dairy products, its effect on body weight may become apparent. We found significant inverse associations between BMI and vitamin D intake in men and women, in both the correlation and regression analyses. The difference was observed in all age groups. When stratifying the data for age, the difference in BMI between those in the upper and lower quartiles of vitamin D intake was 1 kg/m 2 in both subgroups. Vitamin D is important for the intestinal absorption of calcium, and the mechanism whereby a high intake of vitamin D could prevent weight increase should therefore theoretically be identical to that discussed above for calcium. However, this is not the case because we have found opposing effects of calcium and vitamin D intakes, at least in men. It is therefore tempting to speculate that the associations observed do not reflect causal relationships. Like differences in calcium intake, differences in meal patterns between populations may be of importance. In Norway, the main sources of vitamin D are cod liver oil and fat fish rather than exposure to sunlight, and the effect on weight, if causal, could be from a number of substances other than vitamin D. Furthermore, dairy products in Norway are not fortified with vitamin D as in the United States and are therefore a negligible source of vitamin D. Concomitantly, vitamin D intake from exposure to sunlight differs in Norway, and it is to be expected that other associations between vitamin D intake and factors such as BMI exist in our population. The number of subjects included in the study was almost 19,000 and the significant associations detected may not necessarily have relevance in a public health sense. To evaluate this, we calculated the expected BMI if those with a low intake had consumed the RDA dose of vitamin D. This resulted in a decrease in BMI of almost 0.20 kg/m 2 for men and 0.25 kg/m 2 for women in each age group, which could have considerable effect on diseases such as hypertension and diabetes (26). In conclusion, we have found many of the expected relationships between life-style factors and BMI. In addition, our data point out the need for further studies on BMI and calcium and vitamin D intakes. In particular, placebo-controlled intervention studies with calcium and vitamin D, in which the primary end point is weight change, are warranted. LITERATURE CITED 1. Lean, M.E.J. (2002) Pathophysiology of obesity. Proc. Nutr. Soc. 59: 331 336. 2. Taubes, G. (1998) As obesity rates rise, experts struggle to explain why. Science (Washington, DC) 280: 1367 1368. 3. Jacobsen, B. K., Njølstad, I., Thune, I., Wilsgaard, T., Løchen, M-L. & Schirmer, H. (2001) Increase in weight in all birth cohorts in a general population. Arch. Intern. Med. 161: 466 472. 4. Ravussin, E. & Bogardus, C. (2000) Energy balance and weight regulation: genetics versus environment. British J. Nutr. 83: 17S 20S. 5. Davies, K. M., Heaney, R. P., Recker, R. R., Lappe, J. M., Barger-Lux, M., J., Rafferty, K. & Hinders, S. (2000) Calcium intake and body weight. J. Clin. Endocrinol. Metab. 85: 4635 4638. 6. Zemel, M. B., Shi, H., Greer, B., Dirienzo, D. & Zemel, P. C. (2000) Regulation of adiposity by dietary calcium. FASEB J. 14: 1132 1138. 7. Jorde, R. & Bønaa K. H. (2000) Calcium from diary products, vitamin D intake, and blood pressure: the Tromsø study. Am. J. Clin. Nutr. 71: 1530 1535. 8. Altman, D. G. (1999) Practical Statistics for Medical Research, p. 392. Chapman & Hall/CRC, London UK. 9. Prentice, A. (2002) What are the dietary requirements for calcium and vitamin D? Calcif. Tissue Int. 70: 83 88. 10. Smith, S. R., de Jonge, L., Zachwieja, J. J., Roy, H., Nguen, T., Rood, J., Windhauser, M., Volaufova, J. & Bray, G. A. (2000) Concurrent physical activity increases fat oxidation during the shift to a high-fat diet. Am. J. Clin. Nutr. 72: 131 138. 11. Thune, I., Njølstad, I., Løchen, M-L. & Førde, O. H. (1998) Physical activity improves the metabolic risk profiles in men and women. Arch. Intern. Med. 158: 1633 1640. 12. Tavani, A., Negri, E. & La Vecchia, C. (1994) Determinants of body mass index: a study from northern Italy. Int. J. Obes. Relat. Metab. Disord. 18: 497 502. 13. Perkins, K. A., Sexton, J. E. & DiMarco, A. (1996) Acute thermogenic effects of nicotine and alcohol in healthy male and female smokers. Physiol. Behav. 60: 305 309. 14. Tagliabue, A., Terracina, D., Cena, H., Turconi, G., Lanzola, E. & Montomoli, C. (1994) Coffee induced thermogenesis and skin temperature. Int. J. Obes. 18: 537 541. 15. Astrup, A. (2000) Thermogenic drugs as a strategy for treatment of obesity. Endocrine 13: 207 212. 16. Baecke, J. A., Burema, J., Frijters, J. E., Hautvast, J. G. & van der

106 KAMYCHEVA ET AL. Wiel-Wetzels, W. A. (1983) Obesity in young Dutch adults: II, daily life-style and body mass index. Int. J. Obes. 7: 13 24. 17. Merkus, M. P., Mathus-Vliegen, L. M., Broekhoff, C. & Heijnen, A. M. (1995) Extreme obesity: sociodemographic, familial and behavioural correlates in The Netherlands. J. Epidemiol. Community Health 49: 22 27. 18. Teegarden, D., Lin, Y-C., Weaver, C. M., Lyle, R. M. & McCabe, G. P. (1999) Calcium intake relates to change in body weight in young women. (Abstract). FASEB J. 13: A873. 19. Skinner, J., Carruth, B. & Coletta, F. (1999) Does dietary calcium have a role in body fat mass accumulation in young children? Scand. J. Nutr. 34: 45S (abs.). 20. Shi, H., Dirienzo. D. & Zemel, M. B. (2001) Effects of dietary calcium on adipocyte lipid metabolism and body weight regulation in energy-restricted ap2- agouti transgenic mice. FASEB J. 15: 291 293. 21. Hey, H., Lund, B., Sørensen, O. H., Lund, B. & Christensen, M. S. (1979) Impairment of vitamin D and bone metabolism in patients with bypass operation for obesity. Acta Med. Scand. 24: 73S 78S. 22. Jorde, R., Sundsfjord, J., Haug, E. & Bønaa, K. H. (2000) Relation between low calcium intake, parathyroid hormone, and blood pressure. Hypertension 35: 1154 1159. 23. McCarron, D. A., Morris, C. D., Henry, H. J. & Stanton, J. L. (1984) Blood pressure and nutrient intake in the United States. Science (Washington, DC) 224: 1392 1398. 24. Jacobsen, B. K. & Bønaa, K. H. (1990) The reproducibility of dietary data from a self-administered questionnaire. The Tromsø study. Int. J. Epidemiol. 19: 349 353. 25. Cavadini, C., Siega-Riz, A. M. & Popkin, B. M. (2000) US adolescent food intake trends from 1965 to 1996. West. J. Med. 173: 106 114. 26. Wilson, P. W., D Agostino, R. B., Sullivan, L., Parise, H. & Kannel, W. B. (2002) Overweight and obesity as determinants of cardiovascular risk: the Framingham experience. Arch. Intern. Med. 162: 1867 1872.