Soy Product Intake and Hot Flashes in Japanese Women: Results from a Community-based Prospective Study

Abstract

The association between soy product intake and the occurrence of hot flashes was examined in a cohort of 1,106 female residents of Takayama, Gifu, Japan. The women were aged 35–54 years and premenopausal at their entry into the study in 1992. Diet, including intake of soy products and isoflavones, was assessed by means of a validated semiquantitative food frequency questionnaire at study entry. A follow-up mail questionnaire asking about experiences of hot flashes was sent in 1998. During the 6 years of the study period, 101 women had new moderate or severe hot flashes according to the Kupperman test of menopausal distress. After data were controlled for age, total energy intake, and menopausal status, hot flashes were significantly inversely associated with consumption of soy products in terms of both total amount (highest tertile of soy product intake (g/day) vs. lowest: hazard ratio = 0.47; 95% confidence interval: 0.28, 0.79; p for trend = 0.005) and isoflavone intake (highest tertile of isoflavone intake (mg/day) vs. lowest: hazard ratio = 0.42; 95% confidence interval: 0.25, 0.72; p for trend = 0.002). These data suggest that consumption of soy products has a protective effect against hot flashes.

Dietary estrogens are structurally similar to endogenous estrogens. Dietary estrogens mimic the action of endogenous estrogens by binding to estrogen receptors and providing many of the biologic effects observed with endogenous estrogen (1). Therefore, there has been a hypothesis that high consumption of soy products—which are important sources of isoflavone, one of the dietary estrogens—should alleviate the symptoms of hot flashes (2). The low frequency of menopausal symptoms, including hot flashes, among Japanese women has been ascribed to high consumption of soy products (2). Three previous randomized intervention studies showed that soy supplementation reduced the frequency or severity of hot flashes (3–5). However, the subjects in all of these studies were women who were currently experiencing frequent hot flashes. In addition, the duration of intervention was short (12 weeks).

The cumulative effect of usual diet on the onset of hot flashes is of interest. We recently reported a cross-sectional inverse correlation between intake of soy products and severity of hot flashes in a group of community-dwelling women (6). In the present study, we followed up a representative sample of women residing in a Japanese community to see whether soy product intake was prospectively associated with the onset of hot flashes.

MATERIALS AND METHODS

Subjects for this study were drawn from participants in the Takayama Study (7). The Takayama Study commenced in September 1992 in Takayama, Gifu, Japan. Approximately 31,000 persons (92 percent of all residents aged 35 years or more) responded to a self-administered questionnaire that asked for basic demographic data and information on diet, smoking and drinking habits, exercise, and medical and reproductive histories. Dietary history was assessed using a 169-item semiquantitative food frequency questionnaire. Participants were asked to report the average frequency with which each food had been consumed during the previous year and the usual serving size of each food item. We included nine food items for specific soy products (miso soup, tofu, deep-fried tofu, fried bean curd, dried bean curd, natto, houba miso, soy milk, and boiled soybeans). Intakes of soy products and other foods and nutrients were estimated from reported frequencies of intake and portion sizes using standard tables of food composition (8). Detailed information on the questionnaire, including tests of its validity and reproducibility, has been provided elsewhere (7, 9, 10). However, we subsequently revised our method of estimating soy product and isoflavone intake. Besides the above nine items, we took into account some dishes that include soy products as ingredients in order to obtain the estimates. Previously published data on isoflavone concentrations in soy foods were summarized by Wakai et al. (11), and for individual soy foods, we used the median values presented in the review. Total isoflavone intake is the sum of the intakes of daidzein and genistein. Intakes of daidzein and genistein were highly correlated with total isoflavone intake (r = 0.999 and r = 0.995, respectively). In the validity test, Spearman correlation coefficients comparing estimates obtained from this questionnaire and from 12 daily diet records over 1 year were 0.68, 0.63, and 0.65 for soy products expressed as total amount (g), daidzein (mg), and genistein (mg), respectively.

A total of 6,324 women aged 35–54 years reported being premenopausal at the time of entry into the Takayama Study (September 1, 1992). In August 1998, after excluding 222 women who had died (n = 26) or moved to another city (n = 196) by the end of 1996 (based on residential registers for Takayama), we randomly selected 1,500 women who were premenopausal at baseline and sent them a questionnaire asking about their experience of hot flashes. The severity of hot flashes experienced at any time before and during the previous 2 weeks was scored according to the Kupperman test of menopausal distress (12): 0 = no complaints; 1 = mildly unpleasant but not limiting any activity; 2 = moderate, limiting social or other activities but not the ability to work; 3 = severe and incapacitating, causing an inability to work or mix socially. We asked about menopausal status and age at onset of hot flashes. Onset of menopause was defined as age at the last menstrual period prior to stopping menstruation for 12 months. For nonrespondents, we sent the same questionnaire with a reminder letter 1 month after the first one. In total, 1,196 women responded to the questionnaire and 23 women were reported to have died (n = 2) or moved (n = 21) after 1996. Therefore, the response rate was 81.0 percent.

This study was approved by the local institutional review board, and all of the participants provided written informed consent.

Of the 1,196 women who responded to the questionnaire, 47 reported that they had had hot flashes before entry into the study, 11 did not indicate whether they had ever had hot flashes, 7 did not report their age at the onset of hot flashes, and 12 did not report their age at menopause. These women were excluded from the analysis. Furthermore, we excluded the first year of the study period from the analysis in order to minimize the possibility that perimenopausal symptoms at entry into the study might have influenced diet or responses to the questionnaire. Thus, 13 women who reported having had hot flashes during the first year of the study were excluded. Subsequently, the study cohort comprised 1,106 women.

To assess the association between soy product intake and hot flashes, we computed hazard ratios for hot flashes with a severity score of 2 or 3 using Cox's proportional hazards models (13). Women who had scores of 1 were regarded as nonincident cases of hot flashes but were not excluded from the study. Change in menopausal status during the study period was categorized into three groups: premenopausal, postmenopausal for less than 12 months, and postmenopausal for 12 months or more. This time-varying covariate was used for adjustment in the model. Age at study entry, categorized as ≤39, 40–44, 45–49, and 50–54 years, was also included as a covariate in all models. Person-years were calculated for each woman from the date of study entry (September 1, 1992) to the onset of hot flashes or the end of the study period (the date of response to the 1998 questionnaire). After logarithmic transformation, food and nutrient intakes were adjusted for total energy intake using a method proposed by Willett (14). Intakes were then converted into categorical variables based on the tertiles of their distribution in the entire study population. The lowest tertile was used as the reference category. Tests for trend were performed on continuous variables using the median values of the categories. We examined the effects of other potential confounders, such as body size, marital status, education, smoking, parity, exercise, alcohol intake, and dietary variables other than soy products, by including them in the models as covariates. All statistical analyses were performed using SAS programs (15).

RESULTS

Characteristics of the 1,106 women at study entry by tertile of soy product intake are shown in table 1. Person-years of follow-up totaled 6,385. By the end of the study period, 101 women had experienced moderate to severe hot flashes. The hazard ratio for hot flashes was significantly inversely associated with soy product intake in terms of both total amount (g/day) and isoflavone intake (mg/day) after data were controlled for age, total energy intake, and menopausal status (table 2).

Among nondietary factors, years of smoking and number of births were significantly associated with the hazard ratio for hot flashes: For >10 years of smoking versus none, the hazard ratio was 4.3 (95 percent confidence interval (CI): 2.2, 8.2), and for ≥3 births versus nulliparity, the hazard ratio was 0.40 (95 percent CI: 0.18, 0.94). Additional adjustment for these variables did not alter the results substantially: For the highest tertiles of soy product and isoflavone intake versus the lowest, hazard ratios were 0.33 (95 percent CI: 0.16, 0.67) and 0.27 (95 percent CI: 0.13, 0.58), respectively, and p values for trend were 0.002 and 0.0008, respectively. Adjustments for intake of other nutrients, including fat, protein, carotene, crude fiber, and vitamins, did not alter the association between soy product intake and hot flashes.

Thirteen women were using hormone replacement therapy at baseline, and an additional 102 women started using hormone replacement therapy sometime during the study period. Exclusion of these women did not alter the results substantially: For the highest tertiles of soy product and isoflavone intake versus the lowest, hazard ratios were 0.32 (95 percent CI: 0.16, 0.63) and 0.26 (95 percent CI: 0.12, 0.55), respectively, and p values for trend were 0.001 and 0.0003, respectively.

In a subanalysis focusing on the 17 women with severe hot flashes (score = 2), hazard ratios were 0.51 and 0.43 for high versus low tertiles of soy product and isoflavone intake, respectively.

DISCUSSION

The results of the present study suggest that consumption of soy products may protect women against the occurrence of hot flashes. Laboratory data on estrogenic properties of isoflavone plausibly support our finding (16, 17). Our questionnaire was designed to measure an individual's relative intakes of food and nutrients rather than absolute values. The data presented for soy product and isoflavone intake may have been overestimated by the questionnaire, because in the validity study these estimates were about 40 percent higher based on the questionnaire than the estimates based on the diet records. Isoflavone levels in test diets in the previous intervention studies ranged from 76 mg/day to 165 mg/day (3–5). A lower dose of isoflavone in the usual diet may be effective in protecting against hot flashes, considering that the median isoflavone intake in the highest tertile in the present study was lower than the doses used in the interventions. However, intake of soy products seems to be very low in many countries. National per capita consumption data on soybeans obtained from 42 countries indicated that the 75th percentile value was less than 10 percent of that for Japan (18).

The response rate was lower (75.8 percent) among women in the lowest tertile of soy product intake than among women in the second (81.7 percent) and highest (81.7 percent) tertiles. However, the difference was not great, and it is unlikely that women who had had hot flashes were more willing to participate in this study when they had a lower soy product intake.

These results were not due to confounding by other measured dietary or nondietary factors. Psychological stress has been implicated as a trigger for hot flashes. Although we did not obtain information on psychological stress from our subjects, our previous cross-sectional study showed that psychological status, as measured by General Health Questionnaire 12 and the Center for Epidemiologic Studies Depression Scale, had no confounding effect on the association between hot flashes and soy product intake (6).

Subjective reports of hot flashes have been suggested to be a valid indicator of actual occurrence, since they have been found to be reliably associated with increases in skin conductance and finger temperature (19). Because we could not contact our subjects frequently, precise recall of age at onset of hot flashes over the long term may have been difficult. However, it is unlikely that the observed associations resulted from overreporting of hot flashes or reporting of an earlier onset than actual onset time by women with lower soy intakes, because the alleviating effect of soy on hot flashes is not familiar to the public. Recalling whether or not one had had hot flashes during the study period might have been less subject to such a bias than recalling age at the onset of this symptom. When we used a logistic regression model to assess the association between soy product intake and risk of hot flashes without taking into account time of onset, the results were not altered substantially; relative risks were 0.46 (95 percent CI: 0.26, 0.79) and 0.40 (95 percent CI: 0.22, 0.70) for high versus low tertiles of soy products and isoflavone, respectively. Trends were also statistically significant (p = 0.008 and 0.002, respectively).

Since hot flashes generally occur around the time of menopause (20), we included menopausal transition in the models as a time-dependent variable to minimize the possibility that our observations simply reflected the association of diet with age at menopause. At any given time during the study period, we knew retrospectively the number of years prior to menopause for women who underwent menopause before the end of the study, but not for women who remained premenopausal at the end of study. We did not collect data from premenopausal women on signs that might be related to coming menopause, such as irregularity of menstrual cycles and change in the amount of menstrual blood flow. Therefore, it could be argued that there may be a residual confounding effect of menopausal status. However, soy product intake was not associated with age at menopause in the present study. In addition, we focused on 284 women who had undergone menopause during the study period and reanalyzed the data using covariates for menopausal status—i.e., >24 months prior to menopause, ≤24 months prior to menopause, <12 months since menopause, and ≥12 months since menopause. The results were similar to those obtained using all of the subjects (hazard ratios were 0.48 and 0.58 for high versus low tertiles of soy product and isoflavone intake, respectively).

In summary, this prospective study indicated an inverse association of soy product intake with hot flashes. Although traditional hormone replacement therapy is effective in controlling hot flashes, compliance with this therapy is not great, and there is a desire for a more natural approach to the management of menopausal symptoms. Our findings suggest that consumption of soy products is a practical strategy for preventing hot flashes.

This study was supported in part by a grant from the Ministry of Education, Science, Sports, and Culture, Japan.

REFERENCES