The Journal of Nutrition Vol. 128 No. 2 February 1998, pp. 209-213

Decreased Serum Total Cholesterol Concentration Is Associated with High Intake of Soy Products in Japanese Men and Women^1,2

Chisato Nagata³, Naoyoshi Takatsuka, Yoko Kurisu, and Hiroyuki Shimizu

Department of Public Health, Gifu University School of Medicine, Gifu 500, Japan

	ABSTRACT

Abstract Introduction Methods Results Discussion References

The relationship between soy product intake and serum total cholesterol concentration was examined in 1242 men and 3596 women who participated in an annual health check-up program in Takayama City, Japan, provided by the municipality in 1992. The intake of soy products and various foods and nutrients was assessed by a semiquantitative food-frequency questionnaire. Blood samples were collected from fasting subjects to measure the serum total cholesterol concentration. A significant trend (P for trend = 0.0001) was observed for decreasing total cholesterol concentration with an increasing intake of soy products in men after controlling for age, smoking status and intake of total energy, total protein and total fat. This negative trend (P for trend = 0.0001) was also noted in women after controlling for age, menopausal status, body mass index and intake of total energy and vitamin C. An additional adjustment for physical activity, coffee and tea consumption, and intake of cholesterol, carbohydrates, fiber and vitamin E did not change the results. These data suggest a role for soy products in human cholesterol homeostasis.

	INTRODUCTION

Abstract Introduction Methods Results Discussion References

Animal studies have demonstrated that the concentration of cholesterol in blood is lowered by consumption of soy protein rather than animal protein (Carroll and Kurowska 1995, Meeker and Kesten 1941). The effect of dietary soy protein on serum cholesterol concentration has been examined in humans in many clinical trials, but the results have not been consistent. Although significantly decreased plasma or serum cholesterol concentration in some hypercholesterolemic subjects as a result of soy-protein diets was reported in some studies (Bakhit et al. 1994, Sirtori et al. 1977, Verrillo et al. 1985), most studies of normocholesterolemic subjects have shown little difference in effects on plasma or serum cholesterol concentration between soy protein and control diets (Giovannetti et al. 1986, Grundy and Abrams 1983, Meinertz et al. 1988, Mendis and Kumarasunderam 1990, Raaij et al. 1981 and 1982, Sacks et al. 1983). In 1993, the American Heart Association concluded that the positive nature of the effects of soy protein on cholesterol concentration in humans was not so convincing (Chait et al. 1993). However, Anderson et al. (1995) recently performed a meta-analysis of earlier clinical trials and reported that the replacement of animal protein in the diet with soy protein was associated with a significant decrease in serum cholesterol concentration.

To our knowledge, no other cross-sectional studies on the relationship between soy intake and cholesterol concentration currently exist. The relative homogeneity of nutrient intake levels within a population and the errors associated with measurements of diet would make it difficult to detect an association between soy intake and cholesterol level. Soy products such as tofu and miso are traditional Japanese foods and are still popular among Japanese. These characteristics might help us obtain a relatively large variation in intake of these foods among subjects and high reliability in measuring the intake.

Although the previous dietary intervention studies were designed to set up relatively large differences in soy protein intake among the groups compared, the duration of interventions was generally short. The cross-sectional study has limitations in that no causal inference is drawn; however, such a study reflects the association of cholesterol with the usual diet of the subjects over longer periods. Using a cross-sectional study, we examined the relationship between soy products and serum cholesterol concentration in a community in Japan.

	MATERIALS AND METHODS

Abstract Introduction Methods Results Discussion References

The Takayama Study is a prospective cohort study investigating the role of diet and lifestyle on the subsequent development of cancer (Shimizu 1996). About 31,000 persons, 92% of all residents 35 y of age in Takayama City, Japan, participated in the Takayama Study in 1992. Among the participants in the study, 1242 men and 3596 women, who attended the annual health check-up program provided by Takayama Municipality between April and October 1992, were selected as the subjects of this study. We refer to those who were cohort members of the Takayama Study and who did not attend this health check-up program as nonparticipants.

The health check-up program included blood sampling. Blood samples were collected after an overnight fast and serum was obtained by low speed centrifugation at 1500 × g for 10 min. The serum total cholesterol concentration was determined by enzymatic assay (Allain et al. 1974) using an Auto Analyzer (Hitachi, Tokyo, Japan). The reagent used was L-type Wako Cholesterol purchased from Wako Junyaku, Osaka, Japan. The weight and height of each subject were measured and the body mass index (BMI, kg/m²) was calculated.

Detailed information on diet, basic demographic characteristics and medical histories were obtained by a self-administered questionnaire in October, 1992. Dietary history was obtained by a semiquantitative food-frequency questionnaire. Participants were asked to indicate the average frequency of consumption of 169 food items during the year prior to the study and the typical serving size of each food item. There were nine possible response categories for the frequency questions, ranging from almost never, up to six or more times per day. The standard portion size was specified in commonly used units or portions (as judged by a dietitian). Three possible response categories were prepared for the portion size questions: half the standard portion size, standard portion size, and double the standard portion size. We included nine food items for soy products, i.e., miso soup, tofu (bean curd), aburage (deep-fried tofu), ganmodoki/namaage (fried bean curd), koyadofu (dried bean curd), natto (fermented soybeans), houba-miso, soybean milk and boiled soybeans. The individual nutrient and food intake was estimated from the frequency of intake and portion size using the Standard Food Composition Tables published by the Science and Technology Agency of Japan. The intake of soy products was calculated as the sum (in grams) of the nine itemized soy products. We also estimated the intake of soy protein contained in these soy products. Detailed dietary collection methods are more fully described by Shimizu (1996).

Physical activity was assessed from responses to questions concerning the average hours per week spent performing various kinds of activities. The details are described elsewhere (Nagata et al. 1997). Each activity level was assigned an intensity score according to the relative metabolic rate method (Numajiri 1979). The hours spent at each level of activity, weighted by these intensity scores, were totaled to quantify the physical activity level of each individual. The validity of this method was tested (Department of Public Health, Gifu University School of Medicine). The Spearman correlation coefficients comparing energy expenditure estimated from the questionnaire with that measured by the well-established Calorie Counter method (Yanagibori et al. 1991) were 0.69 for both men and women (Suzuki et al., unpublished data).

The protocol of the study complied with the Helsinki Declaration of 1975 as revised in 1983.

The serum total cholesterol and the nutrient intakes were log-transformed for statistical analyses. The nutrient and food intakes were adjusted for total energy by using the method proposed by Willett (1990). The variables for intake of soy products and soy protein were categorized by quartile after adjusting for total energy. The relationships between serum total cholesterol concentration and the intake of soy products or soy protein were assessed primarily by a test for trends using multiple regression techniques (Kirk 1982). Standard regression techniques (Kirk 1982) were also used for analysis of the relationships of selected dietary and anthropometric variables to the intake of soy products or soy protein and to test for adjusted mean values. Correlation analyses (Zar 1984) were used to examine the associations of these variables to serum total cholesterol concentration. The potential confounding factors for the associations of serum total cholesterol concentration with the intake of soy products or soy protein were included in the model simultaneously. All of the statistical analyses were performed using SAS programs (SAS/STAT Version 6.11, SAS Institute, Cary, NC).

	RESULTS

Abstract Introduction Methods Results Discussion References

The validity of this dietary questionnaire has been tested by the use of other dietary assessment methods, one 3-d diet record, four 24-h recalls over 1 y and 12 daily diet records at ~1-mo intervals over 1 y. The Spearman correlation coefficients comparing intake of soy products estimated from the questionnaire with that from 12 daily diet records were as follows: 0.71 and 0.72 for the total amount (grams) of soy products, 0.57 and 0.70 for tofu, 0.83 and 0.76 for miso, 0.66 and 0.79 for natto (fermented soybeans), 0.42 and 0.36 for soybeans and 0.38 and 0.52 for the other soy products in men and women, respectively. The correlation coefficients for intake of fat (total fat, and saturated, polyunsaturated and monounsaturated fat), protein (animal and vegetable protein), cholesterol, carbohydrate, crude fiber, and vitamins C and E from the whole meals between the two methods varied from 0.21 for fat in men to 0.67 for crude fiber in women (the average correlation coefficient for these nutrients was 0.43 and 0.50 for men and women, respectively). Although the correlation for fat intake was low in men (0.21) between the questionnaire and the 12 daily diet records, the correlations comparing the questionnaire with one 3-d diet record and four 24-h recalls were relatively high (0.38 and 0.37, respectively). The average intake of soy products by men and women is shown in Table 1. Soy product intake may be underestimated by our questionnaire. In the validity test, the mean intake of total soy products estimated from the 12 daily diet records over 1 y was ~30% higher than that estimated from the questionnaire.

Table 2 describes the known or potential factors related to serum total cholesterol concentration according to the quartile of soy product intake. All of the nutrient intakes were adjusted for total energy intake. Most of the nutrients listed in the table increased consistently with an increasing intake of soy products in both men and women. Intake of total protein, total fat, saturated and monounsaturated fat, cholesterol, vitamins C and E, coffee consumption and smoking status were significantly associated with both serum total cholesterol concentration and soy protein intake in men. Male smokers had a significantly lower serum total cholesterol concentration (4.64 ± 0.87 vs. 4.82 ± 0.84, P = 0.0001) and a lower intake of soy products (57.7 ± 34.2 vs. 62.2 ± 42.6 g/d, P = 0.02) than nonsmokers. In women, vitamin C intake, BMI, smoking status, coffee consumption and menopausal status were significantly associated with both serum cholesterol and soy product intake. Significantly higher concentrations of serum total cholesterol (5.27 ± 0.87 vs. 4.69 ± 0.81 mmol/L, P = 0.0001) and a higher intake of soy products (55.4 ± 36.6 vs. 45.6 ± 23.7 g/d, P = 0.0001) were observed in postmenopausal women than in premenopausal women. In women, concentrations of serum total cholesterol were 4.95 ± 0.84 and 5.09 ± 0.90 mmol/L in smokers and nonsmokers, respectively (P = 0.02), and soy product intake was 48.2 ± 35.6 and 51.7 ± 31.5 g/d, in smokers and nonsmokers, respectively (P = 0.0001). We also evaluated the relationships between intake of soy protein contained in the soy products and the variables listed in the table. We confirmed that the variables significantly related to both serum total cholesterol and soy protein intake were the same as those significantly related to both serum cholesterol and total amount (grams) of soy products consumed.

Serum total cholesterol concentration decreased with increasing amounts of soy products consumed in men (P for trend = 0.01) (Table 3). This negative trend was strong after controlling for age, smoking status and dietary intakes of total energy, total fat and total protein (P for trend = 0.0001). Crude serum total cholesterol concentration was not significantly associated with soy product intake in women, but a significant negative trend for serum total cholesterol with increasing soy product intake was noted after controlling for menopausal status alone [the means of serum total cholesterol for the lowest to the highest quartiles of soy product intake were 4.96, 4.93, 4.85 and 4.86, P for trend 0.002 (data not shown in Table 3)]. Additional adjustment for age, BMI, coffee consumption, and intake of total energy and vitamin C strengthened the relation (P for trend = 0.0001).

The relationship between the intake of soy protein contained in the soy products and serum total cholesterol concentration was very similar to that observed on the basis of the amount (grams) of soy products consumed.

Additional adjustment for smoking (for women only), alcohol consumption, green tea consumption, physical activity and other nutrient intakes that were significantly or moderately associated with serum cholesterol and the intake of soy products or soy protein did not alter the results. Controlling for each type of fat, saturated, polyunsaturated and monounsaturated fat, instead of total fat did not change the results substantially. Use of fats and total protein as a percentage of energy intake instead of adjusting for total energy intake also did not alter the results.

	DISCUSSION

Abstract Introduction Methods Results Discussion References

The finding in this study of decreasing serum total cholesterol concentration with increasing soy product intake after multivariate adjustments supports the results of some previous clinical trials showing the cholesterol-lowering effect of soy protein (Carroll et al. 1978, Chait et al. 1993). It is of interest that we were able to confirm this relationship in a cross-sectional study. From the results of meta-analysis, Anderson et al. (1995) suggested that 25 g soy protein/d would be associated with a decrease in serum total cholesterol concentration of 0.45 mmol/L. Our dietary questionnaire was designed to measure an individual's relative rather than absolute nutrient intakes. Therefore we cannot directly compare the estimate of soy protein intake in this study with the amount of soy protein fed to the subjects in clinical trials. It is also possible that the relationship between total cholesterol concentration and soy protein intake observed in dietary interventions of short duration might differ from that observed in cross-sectional studies. However, the association observed between soy protein intake and serum total cholesterol concentration, i.e., a 0.31 mmol/L decrease in serum total cholesterol with a 9.6 g increase in soy protein intake between the 1st and the 4th quartiles in men and a 0.23 mmol/L decrease in serum total cholesterol with a 7.9 g increase in soy protein intake between the 1st and the 4th quartiles in women, was not greatly different from that suggested by Anderson et al (1995). A decrease of 0.23 mmol/L in serum total cholesterol was reported by Carroll et al. (1978) among normocholesterolemic women after soy protein was substituted for animal protein in diets for 37 d. No effects of soy protein diets were observed in other studies of normocholesterolemic subjects (Carroll 1991).

It is possible that certain diseases or drugs prescribed for those diseases could influence lifestyle and diet as well as serum total cholesterol concentration. We reanalyzed our data after excluding subjects who had a history of hypertension, diabetes, coronary heart disease, stroke or cancer (260 men and 569 women), but the results were not changed. Exclusion of users of exogenous hormones (53 women) also did not alter the results.

The study population consisted of participants in the health check-up program. This program is provided annually by the municipality to the residents, but the majority attend health check-up programs at their places of employment. It is unlikely that the relationship between soy product intake and serum cholesterol concentration depends on job status. However, because of the voluntary nature of health screening, the issue of selection bias should be considered. Dietary and lifestyle information was obtained from all of the cohort members. Intake of soy products and other nutrients by the participants was generally higher than that by nonparticipants, suggesting that the participants were more health conscious. Selection bias may have occurred if those who had relatively higher intakes of soy products were more likely to attend the health check-up program as a preventive measure and those who had relatively lower intakes of soy products were more likely to attend the program when they had a higher cholesterol concentration than when they had a lower cholesterol concentration. To help reduce the effects of selection bias, we obtained some information on attendance at other screenings (e.g., for stomach, lung, breast and cervical cancer) from each subject and included these variables in the multivariate models. Additional adjustments for these variables did not change the results.

Although the dietary questionnaire has been validated using other dietary methods and moderate-to-good correlation coefficients were obtained for the nutrients of interest, there is a concern that the questionnaire, like other dietary assessment tools, is subject to measurement errors. Although the error is likely to be independent of serum total cholesterol concentration, it is possible that the results are affected when the nutrition variables are controlled in the multivariate models.

We estimated the protein intakes from the consumption of soy products. We did not take into account the soy ingredients such as soy flour and soy protein concentrates used in meat products in the estimate of soy protein intake. However, the major sources of soy protein are the traditional Japanese foods that we selected. Soy flour or soy protein concentrates are rarely used in meat products such as sausage and ham in Japan (information from Kamakura Ham, Yokohama, Japan). Soybean curds are sometimes contained in boiled fish paste (information from Kibun Food, Tokyo, Japan). We estimated the intake of meat and fish products that might include soy ingredients. Daily intakes of sausage/ham and boiled fish paste (mean ± SD) were 4.1 ± 6.4 and 9.2 ± 12.6 g in men and 3.8 ± 6.4 and 8.4 ± 11.2 in women, respectively. Adjustments for these food intakes did not alter the results. We are also interested in the role of whole foods or groups of foods, not only nutrients. The finding of an association between food intake and serum total cholesterol may provide practical guidance for selecting diets to prevent coronary heart disease. The relationship of decreasing serum total cholesterol concentration to increasing soy protein from soy products persisted after adjustment for potentially confounding factors including various dietary components, suggesting the potential of these foods to reduce total cholesterol concentration.

Although we have controlled for several dietary and other factors known or suggested to be associated with serum total cholesterol concentration and soy product intake, we cannot fully exclude the possibility of confounding by other factors. It is also possible that residual confounding effects may have affected the results.

The results of our study add to the existing evidence that suggests that soy protein may be beneficial in human cholesterol homeostasis. Of importance is the fact that we observed the relationship of decreasing serum total cholesterol concentration to increasing soy product intake in a community. Our data will be useful for planning effective education programs to prevent coronary heart disease.

	FOOTNOTES

Manuscript received 22 May 1997. Initial reviews completed 1 July 1997. Revision accepted 31 October 1997.

	LITERATURE CITED

Abstract Introduction Methods Results Discussion References

• Allain C. C., Poon L. S., Chan C.S.G., Richmond W., Fu P. C. Enzymatic determination of total serum cholesterol. Clin. Chem. 1974; 20:470-475 [Abstract]
• Anderson J. W., Johnstone B. M., Cook-Newell M. E. Meta-analysis of the effects of soy protein intake on serum lipids. N. Engl. J. Med. 1995; 333:276-282 [Abstract/Free Full Text]
• Bakhit R. M., Klein B. P., Essex-Sorlie D., Ham J. O., Erdman J. W., Potter S. M. Intake of 25 g of soybean protein with or without soybean fiber alters plasma lipids in men with elevated cholesterol concentrations. J. Nutr. 1994; 124:213-222
• Carroll K. K. Review of clinical studies on cholesterol-lowering response to soy protein. J. Am. Diet. Assoc. 1991; 91:820-827 [Medline]
• Carroll K. K., Giovannetti P. M., Huff M. W., Moase O., Roberts D.C.K., Wolfe B. M. Hypocholesterolemic effect of substituting soybean protein for animal protein in the diet of healthy young women. Am. J. Clin. Nutr. 1978; 31:1312-1321 [Abstract/Free Full Text]
• Carroll K. K., Kurowska E. M. Soy consumption and cholesterol reduction: review of animal and human studies. J. Nutr. 1995; 125:594S-597S
• Chait A., Brunzell J. D., Denke M. A., Eisenberg D., Ernst N. D., Franklin F. A., Ginsberg H., Kotchen T. A., Kuller L., Mullis R. M., Nichaman M. Z., Nicolosi R. J., Schaefer E. J., Stone N. J., Weidman W. H. Rationale of the diet-heart statement of the American Heart Association. Report of the nutrition committee. Circulation 1993; 88:3008-3029
• Giovannetti P. M., Carroll K. H., Wolfe B. M. Constancy of fasting serum cholesterol of healthy young women upon substitution of soy protein isolate for meat and dairy protein in medium and low fat diets. Nutr. Res. 1986; 6:609-618
• Grundy, S. M. & Abrams, J. J. (1983) Comparison of actions of soy protein and casein on metabolism of plasma lipoproteins and cholesterol in humans. Am. J. Clin. Nutr. 1983; 38: 245-252.
• Kirk, D. E. (1982) General linear model approach to ANOVA. In: Experimental Design, 2nd ed., pp. 149-161 and 171-236. Books/Cole, Pacific Grove, CA.
• Meeker D. R., Kesten H. D. Effects of high protein diets on experimental atherosclerosis of rabbits. Arch. Pathol. 1941; 31:147-162
• Meinertz H., Faergeman O., Nilausen K., Chapman M. J., Goldstein S., Laplaud P. M. Effects of soy protein and casein in low cholesterol diets on plasma lipoproteins in normolipidemic subjects. Atherosclerosis 1988; 72:63-70 [Medline]
• Mendis S., Kumarasunderam R. The effect of daily consumption of coconut fat and soya-bean fat on plasma lipids and lipoproteins of young normolipidaemic men. Br. J. Nutr. 1990; 63:547-552 [Medline]
• Nagata C., Kabuto M., Takatsuka N., Shimizu H. Associations of alcohol, height and reproductive factors with serum hormone concentrations in postmenopausal Japanese women. Breast Cancer Res. Treat. 1997; 44:235-241 [Medline]
• Numajiri, K. (1979) Energy metabolism of the physical activity. Tokyo: Research Institute of Labor Science, 62-338.
• Raaij J.M.A., Katan M. B., Hautvast J.G.A.J., Hermus R.J.J. Effects of casein versus soy protein diets on serum, cholesterol and lipoproteins in young healthy volunteers. Am. J. Clin. Nutr. 1981; 34:1261-1271 ^{[Abstract/Free Full Text]}
• Raaij J.M.A., Katan M. B., West C. E., Hautvast J.G.A.J. Influence of diets containing casein, soy isolate and soy concentrate on serum cholesterol and lipoproteins in middle-aged volunteers. Am. J. Clin. Nutr. 1982; 35:925-934 ^{[Abstract/Free Full Text]}
• Sacks F. M., Breslow J. L., Wood P. G., Kass E. H. Lack of an effect of dairy protein (casein) and soy protein on plasma cholesterol of strict vegetarians. An experiment and a critical review. J. Lipid Res. 1983; 24:1012-1020
• Shimizu, H. (1996) A Basic Report on Takayama Study. Department of Public Health, Gifu University School of Medicine, Gifu, Japan.
• Sirtori, C. R., Agradi, E., Conti, F., Mantero, O. & Gatti, E. (1977) Soybean-protein diet in the treatment of type-II hyperlipoproteinaemia. Lancet i: 275-277.
• Verrillo A., De Teresa A., Giarrusso P. C., La Rocca, S. Soybean protein diets in the management of type II hyperlipoproteinaemia. Atherosclerosis 1985; 54:321-331 [Medline]
• Willett, W. (1990) Implication of total energy intake for epidemiologic analysis. In: Nutritional Epidemiology (Willett, W., ed.), pp. 245-271. Oxford University Press, New York, NY.
• Yanagibori R. K., Aoki K., Suzuki Y., Gunji A. Evaluation of methods for measuring daily physical activity in terms of energy expenditure. Jpn. J. Public Health 1991; 38:483-491
• Zar, J. H. (1984) Simple linear correlation. In: Biostatistical Analysis, 2nd ed., pp. 306-323. Prentice-Hall, Englewood Cliffs, NJ.

0022-3166/98 $3.00 ©1998 American Society for Nutritional Sciences