This study utilized published data on the phytoestrogen content of foods to estimate pre-pregnancy phytoestrogen intake among a large, population-based sample of women in the U.S. who recently delivered a baby. These estimations fill an important data gap and are particularly relevant given potential concerns about the effects of phytoestrogen intake on fetal development [12
]. For example, genistein may affect the developing male urogenital tract, including induction of morphologic abnormalities and altered gene expression [12
Previous studies have attempted to quantify dietary phytoestrogen intake. Estimates have tended to rely on data compiled from multiple databases, which used varied measurement techniques and measured different sets of phytoestrogens on different sets of foods. Consequently, even estimates of different phytoestrogens from the same food item may have been derived from different sources, by different methods, and many values may have remained missing. The advantage of the current study is that it could capitalize on recent values generated by a single laboratory, on a wide range of foods common to a Western diet, derived using state-of-the-art measurement techniques on several specific phytoestrogens.
Like our study, de Kleijn et al. examined phytoestrogen intake based on answers to a Willett food frequency questionnaire. Their study population included Caucasian post-menopausal women participating in the Framingham Offspring Study, and they compiled phytoestrogen values from a variety of sources [1
]. In their study, the top contributors to isoflavone intake were beans and peas (26%), tea and coffee (17%), and nuts (15%), versus dairy and fruit in our study. "Soy" contributed only 3% in their study. Their top contributor to lignans was fruit (24%), versus vegetables and fruits in our study. A selected list of vegetables (broccoli, cabbage, cauliflower, lettuce and spinach) accounted for 89% of coumestan intake, versus fruit and dairy in our study. Alcohol made a minimal contribution. As the authors note, their data were most complete for fruits and vegetables, and phytoestrogen estimates were missing for many (if not all) dairy products and for several grain products. This could explain some of the differences with our results. Boker et al. examined phytoestrogen intake among Dutch women (mostly post-menopausal), using a similar approach as de Kleijn et al., but adding phytoestrogen data from some additional sources [31
]. Similar to de Kleijn et al., beans and peas, tea and coffee, and nuts were important sources of isoflavone intake, but so were grain products. Grain products rather than fruits were the main contributors to lignans, and peas/beans were the main contributor to coumestrol.
We observed that total phytoestrogen intake did not differ by education but that intake in the highest quartile was more likely among women who took folic acid-containing vitamin/mineral supplements and among Hispanic women. Milder et al. observed that in the Dutch population, lignan intake was higher among people who were not overweight (BMI < 25), who had higher socioeconomic status, and who did not smoke [32
]. Frankenfeld et al. observed that among post-menopausal women, plasma isoflavone levels were associated with higher intake of fruits and vegetables, lower intake of saturated fats, and higher intake of multivitamin/mineral supplements [33
]. In general, these findings suggest that health behaviors and other nutritional factors, as well as race-ethnicity, may be important correlates of phytoestrogen intake that should be considered in analyses examining associations of phytoestrogen intake with health outcomes.
Some important potential limitations of assessment of phytoestrogen intake in the current study are noteworthy. The NBDPS assessment of coffee and tea intake was restricted to assessing usual intake of caffeinated
coffee or tea before
pregnancy, and is therefore incomplete. Among the sub-set of women who drank these beverages, coffee, and to a lesser extent tea, did make a substantial contribution to phytoestrogen intake. We also did not include phytoestrogen intake from alcoholic beverages in our estimates because information on type of wine and type of liquor, which affect phytoestrogen content, was not available. A total of 37% of mothers drank any alcoholic beverages during the month before pregnancy or the first trimester of pregnancy, but the frequency of alcohol intake tends to be relatively low, with < 3% of mothers drinking daily during the preconception time period (data not shown). Another limitation of the NBDPS food frequency questionnaire is that historically it has not included soy-based items. However, beginning with subjects who had estimated dates of delivery in 2006, the food frequency questionnaire has also included "soy milk or soy yogurt" and "tofu, tempeh or soy burgers." Preliminary data indicate that 13% of women consumed either of these products at least once per month, with only 3-4% of women consuming either food item at least once per week. This low prevalence of consumption suggests that these products are likely to be important sources of phytoestrogen intake for a relatively small proportion of the study population. Additional items that may not be consumed very frequently but are potentially important for a sub-set of individuals due to their high phytoestrogen content were also excluded from the NBDPS questionnaire; e.g., alfalfa sprouts, flax seed, and whole grain varieties of some grain products. Lack of inclusion of these items may contribute to the fact that our reported intake levels are relatively low compared to previous studies, although it should also be noted that in general there is wide variability in reported intakes [1
]. Another potential source of measurement error is that the Kuhnle laboratory focused on food samples obtained locally; it is uncertain whether certain foods would vary between the UK and US, although Kuhnle et al. did not observe clear variability by country of origin for a select group of fruits and vegetables [35
]. Given the largely descriptive nature of the current analysis, we did not explore any formal approaches to address the potential effects of measurement error. Given the retrospective study design, we could not validate our phytoestrogen estimates against a gold standard, such as serum values. However, other studies of phytoestrogen intake derived from food frequency questionnaires have demonstrated good validity when compared with serum or urine values [36