In this study, we developed and evaluated a flavonoid-focused FFQ for Chinese adults. The results showed good reproducibility and relative validity in estimating the intake of six flavonoid subclasses and their 21 primary compounds. The FFQ is the most commonly used tool to assess the intake of dietary flavonoids[32]. It has been widely used in nutritional epidemiological studies on the relationship between flavonoid intake and health. The results of these studies are based on the assumption that the assessment of dietary flavonoid intake is reliable and accurate[32]. Therefore, some challenges must be addressed before an accurate estimation of dietary flavonoid intake can be made. One of the primary challenges is the need for more detailed questionnaires that can accurately cover all the essential sources of flavonoids. Due to the different dietary habits of different regions and countries, residents' food sources of dietary flavonoids in other places also vary. Research on the development of flavonoid-focused FFQs in China is limited. Some studies estimate flavonoid intake solely from soy [33] or fruits and vegetables [34, 35], using a range of 14 to 81 food items. Other studies have evaluated only two flavonoid subclasses (flavonols and flavones) and five primary flavonoid compounds (quercetin, kaempferol, isorhamnetin, apigenin, and luteolin)[36, 37]. The lack of validation and the use of different types of questionnaires in these studies makes it difficult to yield consistent results from the current evidence; therefore, it is difficult to determine a suitable estimated intake of flavonoids for optimal health benefits[38]. Our study designed a flavonoid-focused FFQ containing 147 food items divided into 12 categories, which covered the essential food sources of six flavonoid subclasses in Chinese adults. We believe that this is the first flavonoid-focused FFQ in China that considers the dietary habits of the Chinese population and the primary food sources of the six flavonoid subclasses. The questionnaire provides a detailed analysis of the intake of 21 primary flavonoid compounds and six dietary flavonoid subclasses over one year.
Another primary challenge in validating an FFQ is obtaining a standard reference method and accounting for correlated random errors between the reference method and the FFQ[36]. A systematic review identified the most commonly used reference method for flavonoid-focused FFQs as multiple 24-hour recalls, which are preferred for accurately capturing daily consumption of a varied diet and for their relatively more straightforward administration and analysis than other dietary questionnaires[32]. In this study, the validity and reproducibility of the flavonoid-focused FFQ were assessed by comparing it with an average three-day 24HDR and administering it twice. The flavonoid-focused FFQ developed for the present study showed good validity. Spearman’s and Pearson’s correlation analyses found that almost all flavonoid subclasses and primary flavonoid compounds shared moderate to excellent correlation coefficients (r = 0.284 ~ 0.818). The Spearman correlation coefficient of the original data in our study was higher than those reported in other FFQ studies, which assessed the intake of 5 flavonoid subclasses in Australian adults, including a 62-item food questionnaire (r = 0.100 ~ 0.640)[28], and assessed the intake of 6 flavonoid subclasses in American adults, including a 152-item food questionnaire (r = 0.100 ~ 0.640)[39]. These results are similar to those in studies conducted in China[36, 37], Australia[40], and Japan[41], which assessed the intake of 2 ~ 5 flavonoid subclasses, including 93 ~ 138 food items. The 3 of 20 primary flavonoid compounds with weak or insignificant associations comprised apigenin (flavones), daidzein, and genistein (isoflavonoids) in our study. Previous research also found weak correlation coefficients for these subclasses[28]. This is because only a limited number of food sources contain high amounts of flavones and isoflavonoids. These foods are difficult to capture by 24HDR[21, 42]. Therefore, if the research aims to measure intakes of these subclasses of flavonoids, longer measurement lengths and more detailed dietary assessment tools are necessary. Importantly, assessments of flavonoid intake from FFQ and 24HDR are prone to misreporting bias since they rely on self-reported data. Adjusting for energy intake in the comparisons could help to partially remove this influence[25]. If the variability in the data is more related to misreporting than to energy intake, the energy-adjusted correlation coefficient is reduced[25, 43]. Conversely, if the variability is more related to energy intake, the crude correlations will improve[44, 45]. In our study, the energy-adjusted intakes showed slightly better agreements in the absolute estimated values of almost all primary flavonoid compounds between the two tools but did not change the conclusion of the findings, indicating a limited impact of misreporting on our results. Furthermore, the deattenuated correlations increased because of the correction for the day-to-day variation in intakes.
In relation to variations between individual subjects, we divided 85 subjects into high-, medium- and low-consumption groups according to the intake levels of nutrients, flavonoid subclasses, and primary flavonoid compound intake. Then, the consistency between the flavonoid-focused FFQ and 24HDR was evaluated by comparing the distribution agreement rates of subjects in the three groups. Five of 19 primary flavonoid compounds yielded zero median intakes from the 24HDR, thus making it difficult to compare tertile classification by individual primary flavonoid compound intakes. To account for this, tertile classification by flavonoid subclasses was included. Complete agreement (the same tertile) between the flavonoid-focused FFQ and the 24-HDR ranged from 41.18% (pelargonidin and apigenin) to 77.65% (total flavonoids). The 6 flavonoid subclasses and their 19 primary compounds demonstrated a minor frequency of misclassification and good agreement between the two tools according to Masson’s criteria[46]. Apigenin (flavones), daidzein, and genistein (isoflavonoids) were found to have the weakest correlation coefficients in earlier Spearman’s and Pearson correlation analyses. However, as over half of the participants were classified into the same or adjacent tertile for these primary flavonoid compounds, the flavonoid-focused FFQ still showed good agreement. The assessment of concordance by Kw between the FFQ and the 24HDR displayed a substantial to moderate concordance in their assessment of nutrient and flavonoid subclass intake and a fair to moderate concordance in almost all primary flavonoid compounds; only malvidin, pelargonidin, and apigenin had a weak concordance. In general, these data show that the flavonoid-focused FFQ is useful for assessing intake at this level, but more specific tools may be needed to assess the intake of the individual flavonoids for which the median intake is zero[28]. In our analysis, we found that some primary flavonoid compounds had higher FFQ intake levels than 24HDR. This overreporting is a typical finding when validating an FFQ with many food items[12, 28, 41, 47]. Both the FFQ and the 24HDR have distinct sources of error. The FFQ tends to be impacted by memory bias in the long term, whereas the 24HDR is more susceptible to short-term memory bias. Additionally, there are differences in the perception of portion sizes between the two tools, with the FFQ typically relying on predefined portion sizes[48]. It is also worth noting that the 24HDR method is based on open-ended questions, while the FFQ is designed with close-ended questions in mind. It is possible that the disagreement between FFQ and 24HDR is due to limited food sources of primary flavonoid compounds. Additionally, collecting food consumption data only for a short period (1 week) in this study could make it difficult to capture less frequently consumed food sources in the food diary.
We conducted a reproducibility assessment of the flavonoid-focused FFQ to assess its stability over time. To assess the reproducibility of an FFQ, the best method is to take multiple measurements on a group of participants[49]. In this study, the FFQ was administered twice, with a one-month gap between the two occasions. Our statistical and reproducibility analyses observed a high level of correlation between the two FFQs in estimating nutrients, flavonoid subclasses, and primary flavonoid compounds, except for the primary flavonoid compound pelargonidin, which was slightly overestimated in FFQ compared to FFQ2. Several other studies on FFQ have observed similar results[12, 50]. This could be partly explained by the participants' engagement level and the attention needed to complete the FFQ in full. The ICC and kappa coefficient were calculated to measure variability and consistency by comparing two FFQs. The ICCs ranged from 0.780 to 0.953, and the kappa coefficients ranged from 0.386 to 0.808, suggesting that the flavonoid-focused FFQ has good repeatability and reproducibility. Our results are higher than those obtained in previous studies[51–54].
Using several statistical methods, this study rigorously assessed the reliability and reproducibility of the FFQ. The present study also has limitations that need to be considered. First, this study utilized a flavonoid-focused FFQ consisting of 147 food items, resulting in a more specific and accurate evaluation of dietary flavonoids. However, the increased number of food items needed more time to complete, which could have added pressure to the field investigation. Second, this study was conducted in northeast China, and relevant research conclusions must be verified in other regions and populations in China to determine its applicability. Third, despite good agreement between the FFQ and 24HDR, seasonal variations in some foods may not have been captured accurately by 24HDR due to the length of the validation study (1 year), potentially reflecting differences between the two instruments rather than limitations of the FFQ. Finally, Chinese cuisine incorporates various seasonings, including salt, vinegar, and soy sauce. However, using our tools and reference measures, these condiments were not evaluated for their flavonoid content. The Chinese Food Composition List (6th edition), Phenol Resource database, and NARO database do not provide any information on the flavonoid content of these foods.