食生活の評価方法：セックスとジェンダーの相互作用の分析

The Challenge

Over the past several decades diet-related chronic diseases (DRCDs), including cancers and cardiovascular diseases, increased worldwide (World Health Organization, 2003). Understanding the relationship between dietary factors and DRCDs can lead to cost-effective and science-based disease prevention and management strategies.

Food Frequency Questionnaires (FFQs) are the most commonly used dietary assessment tool in nutritional epidemiological studies because they are cost-effective, easy to administer and are thought to give reasonable estimates of long-term usual intake—see figure below.

Example of a food frequency questionnaire (Modified from Park, 2011, 204)

Individual dietary intakes vary from day-to-day for many reasons, including age, socioeconomic background (food costs and availability), days of the week, food preference and behavioral factors (Beaton et al, 1979, 1983; Gwon et al, 2004; Willet, 2013). Sex and gender are also important diet-determining factors, but these have not been prospectively considered in the development, validation, and implementation of FFQs. Searching PubMed with various combinations of the keywords “FFQ,” “Food Frequency Questionnaire,” “Validation,” and “Validity” yielded 1,124 articles published between January 1983 and May 2014. Excluded were studies conducted in one sex or with subjects other than healthy adults, review articles, articles published in languages other than English, and articles that could not be found through web searches. The resulting 246 validation studies were examined to identify the development process of FFQs used in each study. Among them, only 148 articles described the development process of the FFQ. Since some FFQs were validated more than once, 48 studies that developed FFQs were identified for the remaining validation studies. Therefore, a total of 196 FFQ development studies were analyzed: only 21 studies (10.7%) were found to have considered sex in the development phase of FFQ. The sex of the survey subject was reported more often in determining portion sizes (7.7%) than in selecting food items or “dish” (5.6%). Among the 246 validation studies, only 36.6% reported the results by the sex—see chart below.

Sex of Subject Reported in the Development and Validation Studies of FFQs published between January 1983 and May 2014.

Gendered Innovation 1: Analyzing Sex in Portion Size Increases Accuracy in Dietary Assessment

The Fourth Korean National Health and Nutrition Examination Survey (KNHANES IV) conducted 2007-2009 (Korea Centers for Disease Control and Prevention, 2010) revealed differences between men and women (over age 30 years) along two dimensions important in FFQ development: 1) portion size (influenced by sex), and 2) dish selection (influenced by gender). Analyzing portion size revealed that sex and gender interact: men usually eat larger portions of selected foods, such as steamed rice, a staple of Korean diet. Women, by contrast, chose to eat larger portions of fruits and vegetables, such as oriental melon—see figure below. Dishes consumed frequently also differed by gender (data not shown). These sex and gender differences will impact major sources of energy and nutrients.

Distribution of portion sizes of selected dishes by sex. Men eat larger portions of rice; women eat larger portions of oriental melon. Data from the KNHANES IV 2007-2009 were reanalyzed.

Considering how sex and gender interact is important for identifying the association between diet and cancer. A dish-based FFQ was developed to study diet and cancer associations (Park et al., 2011). One-hundred and twelve dish items were selected considering the amount and frequency of consumption, contributions to energy, nutrients, and cancer-related dietary factors (CRDFs) suggested by the World Cancer Research Foundation (2007). Three portion sizes were assigned for each dish. When the portion size of each dish was re-evaluated by sex, portion sizes of 94 dishes (84%) differed significantly between men and women, 73% of the items were larger in men than in women (Noh et al, unpublished data). These results led to the development of sex-specific portion sizes for all items on the FFQ. A corrected FFQ was developed for each sex.

The original FFQ developed by Park et al. (2011) was validated in a study with 288 adults that consisted of 115 men and 173 women (50% more than men) against 12-day diet records collected over one year (Park et al., 2012). The dataset and the procedures were reexamined after correcting the portion sizes between men and women—while holding dishes available for selection constant.

Method: Analyzing Sex

FFQ validation data can be reevaluated retrospectively to analyze sex

• 1. Data are disaggregated by sex.
• 2. Dish items on FFQ are corrected for portion sizes for each sex.
• 3. Intakes of energy, macronutrients, and CRDFs are corrected for each sex.
• 4. Corrected data are compared to original data to yield a better understanding of dietary intakes.

View General Method

 

Method: Intersectional Approaches

Other factors such as age, BMI, socioeconomic status (SES), and education affect diet and the effects may differ between men and women. Although women are considered more sensitive to social modeling of eating, a comprehensive review of the literature fails to support this hypothesis (Cruwys, et al, 2014). In a study of diet quality and BMI among Canadian adults (Sundararajan, et al, 2014), diet quality was negatively associated with BMI in high BMI group. The author noted that the association was stronger in women. Poorly educated elderly persons in the U.K. showed lower vegetable consumption in men but not in women. Financial hardship, on the other hand, decreased fruit consumption only in women (Conklin et al, 2014). Results of these studies indicate that effects of socioeconomic factors, education, BMI, and exercise on diet are heterogeneous in men and women.

View General Method

 

Gendered Innovation 2: Understanding Gender Differences in Dietary Survey Completion Improves the Quality of Information.

Dietary assessments are conducted based on the assumption that responses to FFQs or any dietary assessment method are similar between men and women. In some cultures, men are less familiar with specific foods and cooking than women. Moreover, women and men’s eating practices, willingness to respond or familiarity with Likert-type scale, bubble format surveys differ. In societies where men do not cook or do not commonly complete questionnaires, they experience difficulties completing in dietary surveys.

Method: Analyzing Gender

• 1. Examine response rates for women and for men in dietary surveys by country.
• 2. Consider gender differences in familiarity with specific foods and cooking.
• 3. Develop FFQs that all participants can understand and answer.

View General Method

 

• 1) p from χ2 test
• 2) Korean Health and Nutrition Examination Survey Ⅳ (2007~2009). Information and data available from here
• 3) NHANES Survey 2007~2010 data combined. Information and data available from here
• 4) National Diet and Nutrition Survey Results from Years 1, 2, 3 and 4 (combined) of the Rolling Programme

  (2008/2009 – 2011/2012); Information and data available from here

FFQs tend to be more difficult or cumbersome to complete than 24-hr recall (24HR) because subjects are asked to answer questions about usual food intakes over a longer period of time and the survey length is commonly over 100 items. Subjects’ responses are also limited to the list of food items provided and serving sizes given on FFQs; despite most FFQs asking open-ended response for “other foods consumed”, few respondents provide additional items in their response.

Subjects who were asked to complete FFQs reported more difficulties answering questions about multiple-item dishes than single dish items, and also more difficulty choosing a portion size than indicating consumption frequencies. In order to address the difficulties in completing FFQs, a qualitative cognitive interview survey was conducted with a FFQ developed by Ahn et al (2007). Among the five men and fifteen women who participated in the study, two men spent about 60 minutes completing the 103-item FFQ. The average length of time to complete the FFQ was 30.6 minutes for total subjects. Of note is the fact that none of the five men cooked at home while all 15 women reported being the main household cook. Several men in the study expressed difficulties understanding the pictures of some foods, and also the description of portion sizes (Lee et al., 2007).

The low response rates of subjects and differential response rates between women and men in dietary surveys (within countries) make it difficult to develop and administer FFQs equally valid for both gender. FFQs need to be designed so that all participants can fill them out with equal accuracy.

Gendered Innovation 3: Gender-specific FFQs Perform Better in Validation and Analysis of Diet-Cancer Relationship

Among the 196 FFQ development studies searched in the literature (see the Challenge section above), only 21(10.7%) developed FFQ with gender consideration. These FFQs are categorized as “Gender-specific FFQ” (GS-FFQ) and all others were categorized as “Not Gender-specific FFQ” (NGS-FFQ). The performances of FFQs of the two categories are examined in the validation studies and nutritional epidemiological studies on diet and colorectal cancer (Lee, 2015).

Method: Rethinking Concepts and Theories

• 1. FFQs are categorized by gender consideration in developmental phase.
  • a. Gender-specific FFQ (GS-FFQ) if developed with gender analysis.
  • b. Not gender-specific FFQ (NGS-FFQ) if developed without gender analysis.
• 2. Validation studies of FFQs in the two categories were identified and intakes of commonly reported nutrients were compared in relation to levels obtained by the reference method.
• 3. Meta-analyses were conducted for prospective cohort studies using GS-FFQs and NGS-FFQs to analyze effects of red meat intake and processed meat intake on colorectal cancer.

View General Method

 

Gender-specific FFQs estimates dietary intake more precisely and with comparable accuracy in men and women. Of the 246 validation studies searched (see the Challenge section), only 45 (18.3%) studies used GS-FFQs. Among these studies, 46.7% reported the results by gender which is considerably higher than 33.3% reporting results by gender in studies using NGS-FFQs. Performances of FFQs in validation studies were examined by comparing intake levels for nutrients commonly reported in both groups—energy, carbohydrate, protein fat, cholesterol, fiber, and calcium. Mean ratios of intakes by FFQ to intakes by the reference method were almost identical for GS-FFQs (0.95 in men and 0.96 in women) but considerably far apart in studies with NGS-FFQs (1.00 in men and 1.12 in women). These results show that the accuracies of nutrient intake estimates are similar in both gender when GS-FFQs are used but significant overestimation of intakes in women may occur when NGS-FFQs are used. (Lee, 2015) The ability of GS-FFQs, compared to NGS-FFQs, to estimate dietary intakes of men and women with similar error have important implications for nutritional epidemiology.

Gender-specific FFQs Perform Better in Detecting the Association of Red Meat Intake and Colorectal Cancer.

Lee (2015) examined the performance of GS-FFQ in detecting diet-disease relation. Of the 297 prospective cohort studies on diet and colorectal cancer selected by the World Cancer Research Foundation (2007, 2011), only 32 studies (10.8%) used GS-FFQs. Meta-analyses were conducted to estimate overall effects of red meat intake and processed meat intake on colorectal cancer. Ten prospective cohort studies that used GS-FFQ (n=5) and NGS-FFQ (n=5) showed an overall effect of red meat intake as a colon cancer risk factor (RR=1.10, 95% CI=1.01 to 1.19). This significance was apparent however only in studies with GS-FFQs (RR=1.16 95% CI=1.00 to 1.32) but not in studies with NGS-FFQ (RR=1.05, 95% CI=0.91 to 1.19)—see figure below.

Meta-analysis of studies reporting association between colorectal cancer and red meat intake selected by WCRF (2007, 2011). Studies in GS group used ‘Gender-specific FFQs’ and studies in NGS group used ‘Not gender- specific FFQs’ as described in Gendered Innovation 3 section.