Introduction
Flavonoids are a large group of naturally occurring plant-based compounds that are commonly consumed through a diet rich in fruit, vegetables, tea, wine and soy-based foods [
1]. Habitual consumption of dietary flavonoids has been consistently linked with improvements in chronic conditions associated with ageing, certain cancers [
2], cardiovascular [
3] and neurodegenerative diseases [
4‐
9]. Flavonoids are divided into six major classes: anthocyanins, flavan-3-ols, flavanones, flavones, flavonols and isoflavones [
10].
Precise estimation of nutrient intake is essential for establishing a relationship between diet and disease. Flavonoids are abundant, wide-spanning and diverse in the human diet, and their quantity in foods is heavily influenced by a food’s growth and processing conditions [
11]. For these reasons, estimations of dietary flavonoid intake need to take into account their complexity and variability. There are substantial variations in population estimates of dietary flavonoid intake [
12‐
17], which may lead to inconsistent associations with health outcomes. A recent review reported a wide range for mean total flavonoid intakes of between 209 to 1017 mg/d (mean 435 mg/d) in Australian, European, and US adult populations [
12]. This variability may relate to true differences in dietary patterns, such as differences in the food supply and cultural eating patterns between countries [
18,
19]. However, it may also reflect well-known limitations associated with the assessment methods typically used to assess flavonoid and subclass intake [
12], described below.
To determine the flavonoid composition of a diet, dietary intake data needs to be cross-referenced with a flavonoid-specific food composition database (FCDB). The dietary assessment method most commonly applied in the literature to determine flavonoid intakes is retrospective analysis of FFQs that aren’t developed or validated to measure flavonoid intakes specifically [
20,
21]. This method, while useful when analysing large existing datasets has limitations for the accurate assessment of flavonoid intakes and can lead to inaccurate results, as often within the FFQs, foods which are nutritionally similar are grouped together (to make the FFQs shorter -i.e., green and red grapes), but these food items often possess very different flavonoid profiles. In addition to the well-known limitations of dietary assessment methods in general [
22], in the case of flavonoids, there are additional methodological issues relating to the choice of FCDBs used to assign flavonoid content information to dietary data. These issues relate to the completeness and appropriateness of the flavonoid FCDB [
12], which is in turn related to availability of analytical food data [
23]. In Australia, for example, very little analytical food data exists for the flavonoid composition of foods, meaning that there are no Australian-specific flavonoid FCDBs to use. Additionally, flavonoid FCDBs are unable to account for inherent variability of the flavonoid composition of foods [
11], which may fluctuate according to cultivar type, season, and/or processing and preparation methods [
24]. Lastly, one recent study demonstrated significant variations in estimates of flavonoid intake when two different flavonoid FCDBs were applied to the same dietary data [
25]. A comparison of the anthocyanin content of fruits and vegetables demonstrated marked variability in anthocyanin content values yielded by three different food composition database sources, namely the USDA tables, Phenol-Explorer and an Australian- specific flavonoid subclass (anthocyanin) database [
23].
Differences in reported flavonoid intakes may be also be attributed to bias associated with different dietary assessment methodologies, leading to further errors in estimation of intake. For example, a Food Frequency Questionnaire (FFQ) and a 24 h diet recall would produce fundamentally different estimates of flavonoid intake, given the inherent differences in the recall and reporting periods of each tool. When assessing flavonoid intake, the majority of studies have applied a FFQ [
26‐
28] to capture habitual intake, while fewer studies have utilised either single [
29,
30] or multiple 24 h recalls [
31], diet history methods [
32] and food records [
33]. The use of FFQs to determine flavonoid intake has limitations, as often a retrospective secondary analysis of flavonoid intake is conducted [
34] from a FFQ tool that has not specifically been designed to assess flavonoid intake. Often, these tools group food items which are nutritionally similar but which possess very different flavonoid profiles. This is especially relevant for assessing fruits and vegetables [
23]. Until recently [
35‐
38], there has been a lack of validated dietary tools for estimating flavonoids and flavonoid subclasses, which is a major limitation to progress in establishment of dietary recommendations.
The known variability associated with estimating dietary flavonoid intake is often attributed to the aforementioned limitations of dietary assessment methodologies, with no consideration of the potential influence of within-individual variation (the inherent day-to-day fluctuation) in flavonoid intake. However, within-individual variation could be significantly contributing to the reported differences in population-based estimates of dietary flavonoid intake. There is substantial within- and between-individual variation for all dietary components, and it is generally well established that macronutrients show smaller variation than micro-nutrients [
39]. Research has established that the number of days of dietary assessment required for accurate estimation of macronutrients intake is a 7-d recording period. However, the majority micronutrients require a longer time period (but less than 1 month) [
40]. It has previously been hypothesised that ‘antioxidant’ dietary components would require more days of dietary assessment than macronutrients, but one study has shown that total flavonoid intakes would require 8 days of dietary assessment, but 10 days would be needed for energy assessment in the same population.
Despite this preliminary analysis, the number of days of dietary data needed to precisely assess flavonoid intakes is currently unclear, with only the one study addressing this issue to date in younger adults and only in relation to total flavonoid and isoflavone intake [
40]. The inherent differences in eating patterns between younger and older adults, underpins the different major dietary sources of flavonoids in these groups, where the contributions of wine and tea to total flavonoid intake increases with age [
41], and given these differences, a focused investigation on the variability in flavonoid intake for older adults is warranted. Information on within- and between-individual variation in flavonoid and subclass intake can be used to calculate the number of days of dietary assessment that are required to precisely estimate intakes of these food components.
Also of potential relevance when considering variability in flavonoid intake assessment, is the potential influence of seasonality. The influence of season on dietary consumption patterns has been established [
42]. Seasonality has been shown to influence nutrient [
43] and antioxidant [
44] intakes, and may influence food availability [
45]. However, the effect of seasonality on dietary flavonoid intake has not yet been adequately investigated. Given that fruits and vegetables are major sources of dietary flavonoids, the effect of seasonality on flavonoid intake could be significant.
The primary aims of the current research were: (1) to assess the between and within-individual variability of dietary flavonoid intake; and (2) to calculate the number of days required to assess usual intake of flavonoids and flavonoid subclasses within a defined level of accuracy using 12 days of weighed food record (WFR) data. A secondary aim of the research was to determine if seasonality impacted on total flavonoid or flavonoid subclass intake in this population.
Discussion
This study shows, for the first time, that precise assessment of total flavonoid intake in older adults requires at least 6 days of weighed food records, and between 6 and 10 days to determine intake of specific flavonoid subclasses with an acceptable degree of accuracy. Season appears to influence intake of subclasses flavanones and flavan-3-ols, but not overall total flavonoid intake.
Substantial within-individual variation and between-individual variation was documented for both total flavonoid intake and intake of flavonoid subclasses in the current study. The within-individual variations ranged from around 80–140% and the between individual variation ranged from around 60–117%, which are both considerably greater than the range suggested for energy and macronutrients. Generally, the expected within- and between-individual variation for energy and other macronutrient intakes is around 25% in free-living subjects [
56]. A number of studies have examined the between and within-individual variability of both macro and micro-nutrient and food intakes [
57‐
60]. An early review by Bingham [
61] identified the mean within-individual CV was lower for energy (23%), and macronutrients (carbohydrate (23%) and protein (27%)). The CV was reported to be greater for vitamins and minerals, such as calcium and iron (34%), ascorbic acid (63%) and retinol (131%). The review concluded that the wider the variation, the greater the number of days required for the reporting period [
61]. It was suggested that 13 days of recording are necessary for 90% of the population to calculate mean energy intake with a standard error of ±10% [
61]. Day-to-day variation in nutrient intake may be the result of an individual’s behaviour [
62], such as differing meal patterns and food availability. For flavonoids, this variability may be attributed to the sporadic nature of consumption patterns of flavonoid-rich foods within the different flavonoid subclasses. For example, red wine or berries are major contributors to anthocyanin intake [
63] but may not be consumed daily. The variation in flavonoid intakes between individuals is also high, with literature showing that sociocultural, economic and ecological factors may be responsible for the variation [
64]. Additionally, small between-person variation may reflect a homogenous population, which does not appear to be the case in the this population, who varied in age and gender [
47]. The within- and between-person variation for some dietary nutrients differs between genders, where women have shown higher CVs than males [
40].
The major sources of variability when determining the flavonoid content of foods are well-known and include the cultivar, growing, processing, and preparation methods, and the variability associated with the analytical methods of flavonoid quantification [
11]. Additionally, differences in a country’s food supply may limit the ability of an international flavonoid FCDB to accurately reflect the flavonoid composition of country-specific foods [
65]. Studies frequently cite these factors as limitations in the interpretation of study findings. However, the potential impact of high within-individual variability on estimates of flavonoid intake has not been addressed. Whilst some studies have averaged repeated measures of flavonoid intake [
66‐
68], so as to minimise the potential impact of within-individual variation, there is no description of the extent of the variability across different time points. The current findings suggest that studies collect as many days of dietary data as possible in order to minimize the effect of within-individual variability on estimates of flavonoid intake. However, increasing the number of days of dietary assessment to minimize this bias is associated with an increase in participant burden and may thereby detract from participant compliance with dietary recording. Therefore, statistically correcting for variability may be more appropriate for large epidemiological studies.
Several statistical methods exist to correct for within-individual variability in dietary intake data [
69]. One method is to collect multiple days of 24-h recall data on each survey participant and average these data [
69]. Another method is to apply a correction factor to the distribution. This method requires estimating the correction factor to be applied, by collecting multiple samples from a representative subset of the survey population for example [
69]. This narrows the population distribution at the extreme ends due to accounting for within-individual variation. More sophisticated statistical modelling methods to account for variability include the Multiple Source Method (MSM) [
70] and National Cancer Institute (NCI) [
71] methods. However, these methods are usually applied to dietary information obtained by repeated short-term instruments, such as a repeated 24-h dietary recalls [
70,
71] in large sample sizes.
Given that flavonoid intake is difficult to quantify, and in the absence of a gold standard approach, methods have been developed for application in various settings, including various techniques within the fields of dietary assessment and biomarker analyses [
65]. A recent review [
65] assessed the available tools to estimate dietary intake of polyphenols, including flavonoids, and identified little consistency across studies when applying FCDBs to estimate intake. Additionally, there is no consensus regarding which dietary assessment tool (e.g. FFQ, 24 h recall, food records etc.) should be utilized to provide the most valid measure of habitual flavonoid intake. However, the use of general FFQs not designed for the purpose of capturing flavonoid intake has been discouraged [
12]. Recently, a flavonoid-specific FFQ for older adults was developed and validated [
38]. Dietary flavonoid intake can also be determined by quantifying relevant biomarkers (e.g. intact phytochemicals or a related metabolite) found in various biological samples. However, there is currently no standardized protocol of how to perform these analyses or which biomarker to target [
12]. Despite the significant problems associated with estimating flavonoid intake using a biomarker (such as within-individual variability in flavonoid metabolism [
12]), future research should focus on the identification of appropriate and easily measurable biomarkers of flavonoid intake. This will be imperative in overcoming limitations associated with the estimation of flavonoid intake using dietary assessment.
There was no statistically significant effect of season on total flavonoid intake in the current study, despite flavonoid intake being relatively high in spring and relatively low in autumn. This finding is not aligned with findings from similar research, which showed that total antioxidant intakes in a Japanese population were highest in winter and lowest in summer [
44]. The authors of this study were able to document differences in participants’ selection of food and beverages across the seasons, and therefore this analysis could be a consideration as a future extension of the current study. The analysis may be crucial to highlight if certain foods are responsible for contributing to the major differences flavanone and flavan-3-ol intakes across seasons. It is possible our lack of seasonal differences for total flavonoid intake reflects, in part, the way in which flavonoid values included in FCDBs are averaged across measurements when determining the flavonoid contents of foods, including different seasons [
11,
24]. As flavonoid-specific FCDBs evolve, information on the influence of seasonality on the flavonoid content of foods may become more widely available. A limitation of this analysis is that the WFR data was collected across three seasons for each participant only. Ideally, dietary information would be collected mid-season, and in all seasons for each individual in future research.
The sample used for the current analysis was originally collected for a validation study of a FFQ developed for a prospective cohort study. The burden to participants entailed in the collection of twelve days of weighed food records is substantial and the sample size, while typical of validation studies of this nature, was relatively small. The reason for utilizing this dataset to estimate flavonoid intake in older adults relates to the richness of the dietary data. WFRs are likely to provide a more accurate estimation of flavonoid intake in comparison to other dietary assessment methods, such as repeated 24 h recalls. The dietary data collected in the total BMES sample was a FFQ, which grouped nutritionally similar foods (e.g., apples and pears). Given that such foods have significantly different flavonoid profiles, however, the FFQ may be unsuitable for accurately estimating flavonoid intake. Additionally, several major flavonoid contributing foods were not included in the BMES FFQ. Thus, despite the relatively small sample size, the depth of the dietary data from the WFRs in this group is a major strength when estimating flavonoid intake. We have previously compared flavonoid intake in this population to other national and international estimates for older adults, showing that older adults tend to consume higher amounts of dietary flavonoids when compared with younger age groups [
17]. This may be related to higher intakes of tea and wine as people age [
17]. It is difficult to compare flavonoid intakes in older adults across populations and studies because of differences in dietary assessment methods and the use of different FCDBs. Nevertheless, flavonoid intakes are reported to range from around 21.2 mg/day to 191.2 mg/day in this population [
72].
An additional limitation of the current study is that the data used for this analysis was collected in the 1990s. However, in the BMES population, fruit and vegetable consumption did not significantly change from baseline to the 10-year follow up [
73]. Some changes in dietary patterns related to fat (MUFA, PUFA, SFA) and total sugar (not CHO) intake [
73] may have occurred during this period, but these macronutrients are not generally associated with flavonoid-rich foods. Nevertheless, the generalizability of the study findings may be limited by the changing food supply. Despite the age of the comprehensive dietary data used by this study. The USDA database chosen as our reference flavonoid FCDB was comprehensive enough to assign the WFR food items flavonoid content values. However, the validity of using a current (present-day) flavonoid FCDB to retrospectively assign flavonoid contents to foods collected approximately two decades earlier is uncertain, and these methodological limitations should be considered when interpreting the findings of this research. This study utilised an international FCDB, and therefore the potential inaccuracy of the flavonoid content of foods for Australian produce is a limitation of this study. The USDA [
1] recognises that flavonoid contents in foods are influenced by cultivar types, and the growth and processing conditions of foods, but this is an issue across all flavonoid FCDBs. Therefore the USDA database was an appropriate choice for this study as it is a comprehensive resource and is commonly applied across studies. However, improvements in country-specific flavonoid FCDBs, ideally integrated into existing dietary analysis software, are vital to improve the accuracy and ease of flavonoid intake estimates in future studies.
Lastly, the sample size of the current study did not permit stratification of findings by gender, which is another limitation of the analysis. We have previously reported a significant difference in energy intake between men and women in the current study population [
17] but a gender difference was not evident for flavonoid intake [
17]. The vast majority of flavonoids are provided by tea, a low energy food, such that accounting for differences in energy intake is unlikely to uncover sex differences in flavonoid intake. Further research is needed to investigate the influence of energy intake, gender, or other confounders for diet, such as age and levels of physical activity on variations in flavonoid intake.
In conclusion, further research is needed to identify the determinants of the day-to-day variation in flavonoid and subclass intake within and between individuals, and whether a high variability in flavonoid intake has any biological implications in terms of metabolism, uptake and excretion [
62]. Additionally, given the limitations of our study, further research is required to confirm our findings and to determine the appropriate number of days to accurately determine flavonoid intake. Comprehensive, Australian-specific flavonoid FCDBs are also needed, ensuring flavonoid content-values are representative across all seasons. Our study has shown that the within- and between-individual variation in flavonoid intake is considerable and needs to be accounted for in dietary assessment methodology. Additionally, the collection of dietary data in different seasons may not significantly influence estimates of total flavonoid intake but may influence the reported intakes for flavanones and flavan-3-ols. The findings of this study suggest that at least 6 days of weighed food records for total flavonoid intake, and up to 10 days for individual flavonoid subclasses, should be collected to reduce the bias associated with within-individual variations in intake.