This is the first study to assess the suitability of five potential whole grain surrogate estimates to measure total whole grain intake in a Western population within the context of epidemiological research. In addition, whole grain intake based on the Healthgrain Forum’s whole grain food definition was examined in relation to total whole grain intake. Among 5094 Finnish adults participating in the national FinHealth 2017 Study, the correspondence was consistently the greatest between total whole grain intake and intakes of whole grain (whole grain food definition) and the combination of rye, oat and barley regarding both correlations and cross-classification of quintiles. Consumption of rye bread and rye corresponded reasonably well with total whole grain intake.

Thus far, observational studies have focused on the accuracy of whole grain intake estimates when whole grain food definitions with cut-off values for whole grain content have been applied. In the UK, applying a 51% cut-off resulted in a significant reduction in the estimated whole grain intake compared to total whole grain intake among 3073 British participants aged ≥ 1.5 years [13]. Another study comparing two cohorts of British adults (n = 2086, aged 16–64 years; n = 1692, aged 19–64 years) reported that the 51% cut-off underestimated whole grain intake by up to 27% in comparison with a 10% cut-off [19]. With a 25% cut-off, the underestimation was up to 10%. Moreover, when the Healthgrain Forum whole grain food definition was applied in the Australian (n = 12 153, aged ≥ 2 years) and Swedish (n = 1797, aged 18–80 years) populations, significantly smaller whole grain intakes were detected compared to total whole grain intake [21, 34].

In our study population, whole grain intake was 7–8% lower when the Healthgrain Forum definition was applied compared to total whole grain intake. However, unlike the previous studies, we assessed dietary intake utilizing the FFQ instead of a 24-h recall or a food diary. In our epidemiological context, whole grain food definition-based whole grain intake corresponded extremely well with total whole grain intake, the correlation being 0.99 and the proportion of participants categorized into the same or adjacent quintile in accordance with total whole grain intake being nearly 90%. This is in accordance with our previous findings utilizing 24-h dietary recalls in a subsample of the same study population, suggesting that whole grains are mainly consumed as foods with very high whole grain content, such as rye bread and porridge [35].

Along with the definition-based whole grain intake, all rye-based variables corresponded well with total whole grain intake. The correspondence was especially good regarding the combination of rye, oat and barley, with a correlation of 0.99 and the proportion of participants categorized into the same or adjacent quintile being 100%. Overall, a good correspondence between rye-based variables and total whole grain intake was expected, as in Finland, rye and rye bread are the primary cereal and food sources of whole grains [35]. Moreover, the results regarding the combination of rye, oat and barley were in accordance with our expectations as oat is the second-most important cereal source for whole grains among Finnish adults, and in Finland, both barley and oat are predominantly consumed as whole grain [35].

Of the examined variables, dietary fiber intake and bread consumption had the weakest correspondence with total whole grain intake. This was not surprising, as dietary fiber can originate from numerous other food sources than whole grain cereals in the diet. Furthermore, alongside rye bread, wheat-based bread with low or no whole grain content has been found to cover a significant proportion of bread consumption in our study population [26].

In cross-classification, dietary fiber intake and bread consumption were the most noticeably affected by the exclusion of energy under-reporters in their correspondence with total whole grain intake. The exclusion decreased the proportion of participants categorized accordingly in the lowest and lowest two quintiles and increased the proportion of grossly misclassified. This suggests that the participants within the lowest whole grain intake quintiles under-reported their intake of fiber sources other than whole grain-containing foods. These may include, for example, wheat-based bread and confectionary with no whole grain ingredients. This is supported by the similar but less drastic phenomenon in bread consumption in relation to total whole grain. As the effect of excluding energy under-reporters was the greatest regarding the proportion categorized accordingly in the lowest quintiles, and only minor changes appeared in the overall classification of participants in the same quintiles, energy under-reporting seemed to occur mainly in the lower end of dietary fiber intake, bread consumption and total whole grain intake. As energy under-reporting is a significant challenge in self-reported dietary intake data in most populations, sensitivity to under-reporting should be considered in choosing a suitable surrogate estimate. Indications of the crudeness of dietary fiber as a surrogate estimate for total whole grain intake have also appeared in other populations. For example, in the USA, only 15% of fiber intake was estimated to originate from whole grain-containing foods [36].

In the subgroup analysis, dietary fiber intake and bread consumption appeared to have the biggest between-group differences in the correlations with total whole grain intake. Dietary fiber intake and bread consumption correlated better with total whole grain intake in men than in women and regarding bread consumption, better in participants with obesity than participants with overweight or normal weight. The differences between sexes are in accordance with earlier findings in Finnish adults, suggesting that the contribution of cereals on fiber intake is more pronounced among men, while women get fiber more variedly from different sources [26]. Correspondingly, in men, bread consumption contributes more to total whole grain intake than in women [35]. The stronger correlation in participants with BMI ≥ 30 kg/m2 compared to others indicates similarly a more pronounced role of bread as a whole grain source. However, overall, the differences in correlations between sex, age, educational level and BMI groups were small and irrelevant to the interpretation of the results.

The strongest correspondences with total whole grain intake were mediated by rye intake, including especially the combination of rye, oat and barley, as well as rye and rye bread. As high rye consumption is a distinctive characteristic of the traditional Finnish (and some other Nordic) diet, our results may not be applicable to other populations with diverging whole grain sources. Differing whole grain sources also challenge the suitability of whole grain food definitions in the estimation of whole grain intake in different populations. The major whole grain sources within a population define how much the applied cut-off for whole grain content in foods underestimates actual whole grain intake. Thus, even with the same cut-off, the level of underestimation may drastically differ between populations. Estimating whole grain intake based on foods with required whole grain content instead of any amount of whole grains seems to be especially problematic when a large proportion of whole grains are consumed from foods with low whole grain content, as is displayed in the previous studies [13, 19]. Applying such an estimate in epidemiological research might distort the display of existing associations between whole grain intake and health outcomes.

In our study population, applying the Healthgrain Forum definition with a 30% cut-off resulted in only minor reductions in the estimated whole grain intake. Even a stricter cut-off value of 51% would unlikely result in a much greater reduction as only three foods within the FFQ items with small relevance for total whole grain intake would be further excluded from the estimation. Moreover, the participants were predominantly organized into quintiles accordingly between total whole grain intake and definition-based whole grain intake. This suggests that in the Finnish population with a high proportion of total whole grain intake originating from foods with high whole grain content, whole grain intake estimated based on the Healthgrain Forum definition would be a good estimate for an epidemiological study. Furthermore, utilizing the food-level estimate might provide useful insights into whole grains’ health associations while considering the complex interactions between food components. Findings in a few previous studies examining definition-based whole grain intake estimates regarding specific health outcomes have also indicated that the food-based estimate would be suitable for epidemiological research. In a study conducted in the Australian and Swedish populations, applying the Healthgrain Forum definition had only minor effects on the associations between whole grain intake and risk factors of cardiovascular diseases (CVD) [34]. Furthermore, in an American prospective cohort of 42 850 males aged 40–75 years, the association with CVD incidence did not differ whether total whole grain intake or whole grain intake based on a 51% cut-off was used [37].

Overall, further research on potential surrogate estimates of total whole grain intake in different populations is needed. By establishing the suitability of surrogate estimates, research on whole grain intake can be facilitated in situations where total whole grain intake cannot be calculated. Furthermore, standardized use of surrogate estimates validated in the target population promotes the comparability and consistency of the findings and may further consolidate the evidence on whole grains’ health associations. For the use of epidemiological studies, demonstrating the suitability of surrogate estimates regarding specific health outcomes is as well called for. Moreover, further efforts are required to standardize whole grain definitions in research as, along with the inconsistencies in estimation methods, inconsistent definitions limit the comparability between studies and add to the discrepancies in results.

Strengths of this study include a comprehensive evaluation of the correspondence of food, ingredient, and nutrient level variables with total whole grain intake. As the availability of dietary variables differs in research settings, examining the applicability of different level variables facilitates finding a suitable surrogate estimate in varying situations. Furthermore, we conducted the analyses in a large, population-based sample of the Finnish adult population. Dietary intake was assessed by utilizing an FFQ that has been repeatedly validated in the Finnish adult population. Total whole grain intake was estimated by applying a comprehensive, newly developed whole grain database on a dry-weight basis, covering whole grain intake from the whole diet. To consider the possible misreporting of food consumption arising from self-reported dietary intake, we conducted the analyses both in the total population and excluding energy under-reporters.

The limitations of our whole grain intake estimation are linked to the general challenges in utilizing food composition databases. The information on whole grain content in foods originates from different sources, including ingredient labels, information from manufacturers, whole grain content in similar food products and estimated recipes. Thus, the accuracy of the information may vary. Furthermore, due to the constantly expanding and changing supply of industrial food products, the food composition database may not entirely reflect the current food supply at the data-gathering moment. However, these challenges are not unique to our study but concern more widely the application of food composition databases. In utilizing the Healthgrain Forum definition for whole grain intake estimation, we excluded one compliant food within one FFQ item to facilitate the calculation process. This may have affected our findings on the whole grain food definition-based whole grain intake. However, as the weight of the food in question on the FFQ item was very small, it is unlikely that including it in the estimation would have changed the results significantly. Especially as the effect would have been toward even stronger correspondence between total whole grain and definition-based whole grain intake. Finally, as already mentioned, our results cannot necessarily be applied to other populations due to differing characteristics of food consumption and whole grain sources between populations. However, our findings may provide context and a basis for comparison for future studies, furthering the validation of suitable whole grain surrogate estimates in different populations. Moreover, our results strengthen the findings of previous epidemiological studies that have examined whole grain intake in Finnish adults utilizing some of the surrogate estimates we found acceptable.

In conclusion, our study suggests that rye-based variables, especially the combination of rye, oat, and barley, are suitable surrogate estimates for total whole grain intake in the context of epidemiological research in the Finnish adult population. Similarly, whole grain intake based on the Healthgrain Forum whole grain food definition corresponded well with total whole grain intake. Conversely, dietary fiber intake and bread consumption appeared less suitable estimates for total whole grain intake due to their weaker correspondence with total whole grain intake and sensitivity to energy under-reporting. With more harmonized and validated methods to estimate whole grain intake, the evidence on whole grains’ health associations could be consolidated and the efforts to increase whole grain intake in the population strengthened.