Reproducibility and validity of a 45 item food-frequency questionnaire for inuit in Greenland

ABSTRACT

Since 1993, dietary assessment has been carried out in Greenland as part of recurrent population health surveys. In preparation for the next survey in 2024, 91 participants from the survey in 2018 were selected for a validation study of the Food Frequency Questionnaire (FFQ). The 91 participants were reinterviewed 38–50 months after the first FFQ and invited to complete a food diary. As part of the 2018 survey, blood was analysed for mercury. The food diary was completed by 65 participants. The agreement between the two FFQ interviews was very good for macronutrients and fatty acids (p > 0.20), whereas the calculated intake of mercury was 22% higher in the second FFQ (p = 0.04) due to a higher intake of whale meat and muktuk (whale skin). The agreement between the second FFQ and the food diary was good for local food, imported meat and cakes/sweets/snacks but fruit and vegetables, dairy products, beverages and added sugar were significantly underreported in the food diary. Food items not included in the FFQ were identified from the food diaries. The correlation between the intake of marine mammals and blood mercury was moderate (Spearman’s rho = 0.41–0.50; p < 0.0001). The results will inspire future dietary studies in the circumpolar North.

Introduction

A food-frequency questionnaire (FFQ) is a convenient method to assess dietary intake in epidemiological studies [1]. There are numerous internationally validated FFQs, but dietary patterns vary among countries and ethnic groups, and it is therefore usually necessary to develop locally specific FFQs. The Inuit, for example, subsist on a diet rich in locally harvested marine mammals and fish in addition to imported food [2] and it is therefore needed to add food like seal meat, whale meat, caribou and game birds to FFQs among the Inuit. Furthermore, large population health interview studies usually cover multiple topics and there is a continuous need to reduce the number of questions. In Greenland, a 66-item FFQ with portion sizes and seasonal variation of locally harvested food was developed for the Inuit Health in Transition study in 2005–2010 [3]. This FFQ was considered to be too lengthy for a general health survey and was subsequently reduced to an FFQ with 45 food items with portion sizes but not seasonal variation [4]. [5] This is now considered the standard FFQ in Greenland. An earlier version of the FFQ was validated by 24- and 48-h dietary interviews on traditional food and by blood mercury analyses with good agreement between the methods [6]. The 66-item FFQ of the study in 2005–2010 was validated for the consumption of locally harvested food. Whole blood mercury was found to be a better biomarker for the consumption of marine food than eicosapentaenoic acid (EPA), docosahexaenoic acid (DHA) and total N-3 fatty acids in erythrocyte membranes [7].

The purpose of the present study was to evaluate the reproducibility of the revised 45-item FFQ among the Inuit in Greenland, to validate it by comparison with a food diary and to compare its questions about locally harvested food with whole blood mercury as a biomarker.

Material and methods

Data collection

Data was collected from 2016 to 2019 as part of a countrywide survey of health and risk factors for health [4]. The participants were selected through a stratified random sample of adults in Greenland, who had been born in Greenland or Denmark. From each of five strata defined according to geographical criteria (South, Mid, Northwest, North and East Greenland), 20 towns and villages were chosen at random. From 12 towns, a random sample of inhabitants aged 15+ years were invited and from 8 randomly selected villages all inhabitants were invited to participate in the study. Data was collected by interviews, self-administered questionnaires, clinical examinations and blood sampling. The participation rate was 52%. Questionnaires were developed in the Danish language, translated into Greenlandic (Kalaallisut), back translated and revised. Interviews were conducted in the participant’s language of choice, which was most often Greenlandic, by native Greenlandic speaking interviewers who had been trained in the study procedures. A total of 2436 Inuit participated in the survey. Inuit ethnicity was defined by the interviewers at enrolment based on primary language and self-identification. In a 45-item food frequency questionnaire, participants were interviewed about how often they had consumed each food item within the last 12 months and their usual portion size (FFQ Interview 1). The questions were, for example : How often do you eat seal meat? When you eat seal, whale and muktuk, how much do you eat at a time (portion size)? Food frequency was recorded as the number of times per day/week/month/year the food item was consumed. Portion sizes were estimated from photos of four serving sizes presented to the participants by the interviewer (Figure 1); fractions of the portions and added portion sizes were allowed.

Figure 1.

Four portion sizes of fish: 300 g, 200 g, 100 g and 50 g.

Data management

Frequency was recoded into times per month. Portion sizes were recoded into gram and for each food item the daily amount consumed was calculated as g per day (frequency × portion size/30.4). Missing information about frequency was coded as 0 and missing information about portion size was coded as the population median separately for men and women. The estimated intake of macronutrients, fatty acids and mercury was calculated from the daily consumption multiplied by published values for protein, fat, carbohydrates, saturated and unsaturated fatty acids and mercury [8]; [9]; [10] [11,12]. Cooking loss was included in the calculations as appropriate [13,14]. Total energy was calculated as protein × 17 kJ/g + fat × 37 kJ/g + available carbohydrate × 17 kJ/g + dietary fibre × 8 kJ/g. Alcohol was not part of the FFQ but was recorded elsewhere. Alcohol consumption was seriously underreported [15] and only 76% of participants with valid information on diet had information on their use of alcohol. Alcohol was therefore not included in the calculation of total energy.

Underreporting and overreporting of total energy intake was handled by left-truncating the distribution at 2100 kJ per day for women and 3350 kJ per day for men [1] and right-truncating the distribution at the 97.5% percentile (19,000 kJ for women and 22,000 kJ for men). The recorded energy intake was outside these limits for 4.6% of the participants bringing the number of Greenlandic participants with an acceptable energy intake to 2325. The estimated total energy expenditure was calculated as Resting Energy Expenditure (REE) multiplied by an estimated Physical Activity Level (PAL) [16].

Validation study

One hundred Inuit participants were selected for the validation study from the capital (Nuuk), one medium-sized town (Aasiaat) and one village (Attu). In 2022, the FFQ was repeated among these (FFQ Interview 2). Food frequency data was treated in the same way as described above. All 100 participants had recorded energy intakes in the second survey that were within the acceptable limits but due unforeseen events during the recruitment of participants in the village only 91 had an acceptable energy intake in FFQ Interview 1.

Food diary

The 100 selected participants were asked to complete a food diary, preferably daily over 3 weeks on three different seasons but if this was not feasible then for as many days as possible. The information in the food diaries was translated into Danish and manually recoded into the 45 food items that were part of the FFQ plus 10 novel food items. For comparisons between FFQ and food diaries, the food items were categorised into 11 food groups. A total of 65 participants recorded 4296 food items for 452 days. On average, participants recorded their diet on 7.0 days (median 3.0; range 1–46) with 9.5 food items per day. Participants were requested to record the number of food items and their portion sizes. However, the recording of portion sizes was often imperfect and for 34% of recorded food items no portion size was noted at all. Accordingly, comparisons of FFQ and food diaries were only performed for food frequencies.

Biomarkers

Whole blood mercury was analysed at Centre de toxicologie du Québec/INSPQ [17] by ICP-MS (Inductively coupled plasma mass spectrometry) with a detection limit of 0.5 nmol/Litre. Information was available for 1719 Inuit participants.

Statistical methods

Statistical analyses were carried out with IBM SPSS Statistics version 28.01. In Tables 1–3, outcome variables are naturally log transformed in order to approach normal distributions and means were back transformed for presentation as geometric means. In Tables 5 and 6, Spearman’s correlation was used on untransformed values. The Bland-Altman plot in Figure 2 was drawn in Excel Microsoft 365 using untransformed values. Other statistical tests included two sample T-test, Anova, chi-square test and Kappa statistics.

Table 1.

Participant and survey characteristics. Survey participants with recorded energy intake within the acceptable limits.

	Health survey 2018 FFQ Interview 1		Follow-up survey 2022 FFQ Interview 2			Food diary 2022
	n = 2234 not selected	n = 91 selected for follow-up			n = 65 with food diary
Age (median)(range)*	49 (15–94)	46 (23–75)	49 (26–78)		51 (24–79)	51 (24–79)
	p = 0.54
Sex (% women)**	55.6	58.2	58.2		63.1	63.1
	p = 0.67
Interview date (median)*	May 2018	Jan 2019	Mar 2022		Mar 2022	Aug 2022
Place of residence**	%	%	%		%	%
Capital	16.7	39.6	39.6		30.8	30.8
Towns >2000 pop	31.2	36.3	35.2		33.8	33.8
Smaller towns, villages	52.1	24.2	25.3		35.4	35.4
	p < 0.001
Interview season**
Winter	47.4	81.3	46.2		46.2	39.1
Spring	11.2	0.0	19.8		21.5	37.5
Summer	9.4	0.0	5.5		13.8	10.2
Autumn	32.0	18.7	28.6		18.5	13.2
	p < 0.001 p < 0.001				p = 0.29
	kJ/day	kJ/day	kJ/day
Recorded energy intake***	8162	7655	7497		-	-
	p = 0.13 p = 0.61

*Anova.

**Chi-square.

***Geometric mean; Anova.

Table 2.

Calculated intake of energy, macronutrients and dietary groups according to two Food Frequency Interviews; 91 cases with acceptable energy intake in both surveys. Geometric means.

	FFQ Interview 1	FFQ Interview 2	Difference (%)
Daily consumption	n = 91	n = 91		p*
Energy (kJ)	7655	7497	−2.1	0.61
Fat (g)	53.3	54.0	1.3	0.82
Protein (g)	226.9	219.4	−3.3	0.46
Carbohydrate (g)	87.4	83.1	−5.0	0.31
Refined sugar (g)	51.6	57.2	10.9	0.52
Dietary fibres (g)	22.6	21.1	−6.8	0.24
Polyunsaturated fatty acids (PUFA) (g)	7.4	7.3	−0.4	0.94
Eicosapentaenoic acid (EPA) (g)	0.4	0.4	2.6	0.74
Docosahexaenoic acid (DHA) (g)	0.5	0.5	−1.4	0.87
Mercury (microg)	12.3	15.0	21.8	0.04

p* two-sample T-test.

Table 3.

Food frequencies in FFQ Interview 2 and the food diary. Meals per month and difference between the two methods. N = 65. Geometric means.

Food group	FFQ meals per month	Diary meals per month	Difference %	p*
Marine mammals	5.2	6.2	−16	0.39
Fish	5.2	4.0	31	0.23
Other locally harvested food	9.8	10.0	−2	0.93
Imported meat	29.7	29.5	0	0.97
Fruit and vegetables	57.4	25.3	127	<0.0001
Dairy products	30.2	16.0	89	0.0002
Bread, pasta, rice	63.1	50.4	25	0.0009
Cakes, sweets	10.2	9.1	11	0.61
Beverages	22.7	9.2	145	0.0004
Junk food, snacks	6.5	4.4	48	0.03
Coffee/tea with sugar	14.7	4.8	206	0.0006
All	370.6	268.0	38	<0.0001

p* two sample T-test.

Table 5.

Spearman’s correlations of food intake with whole blood mercury. N = 1719.

	Men (n = 736)		Women (n = 983)
	ρ	p	ρ	p
Frequency of seal meat	0.46	<0.0001	0.39	<0.0001
Frequency of marine mammals	0.50	<0.0001	0.41	<0.0001
Frequency of fish	0.30	<0.0001	0.22	<0.0001
Frequency of kalaalimernit	0.44	<0.0001	0.35	<0.0001
Seal meat (E%)	0.46	<0.0001	0.35	<0.0001
Marine mammals (E%)	0.50	<0.0001	0.38	<0.0001
Fish (E%)	0.27	<0.0001	0.23	<0.0001
Kalaalimernit (E%)	0.44	<0.0001	0.34	<0.0001
Calculated intake of mercury (µg/day)	0.49	<0.0001	0.38	<0.0001

Table 6.

Spearman’s correlations of frequency of consumption of marine mammals with whole blood mercury according to urbanisation. N = 1719.

	Men (n = 736)		Women (n = 983)
	ρ	p	ρ	p
Capital (Nuuk) (17,800 pop)	0.47	<0.0001	0.63	<0.0001
Communities >2500 pop	0.48	<0.0001	0.39	<0.0001
Communities 500–2499 pop	0.42	<0.0001	0.23	<0.0001
Communities <500 pop	0.26	0.004	0.27	<0.0001

Figure 2.

Bland-Altman plot of energy measured in two FFQ surveys 41 months apart. N = 91. Dotted lines are 95% confidence intervals.

Results

There were some differences in metadata for FFQ Interview 1 between the 91 participants who participated in the validation study and those who did not (Table 1). The 91 participants selected for the validation study were slightly younger than those not selected and more were women, but the differences were not statistically significant (Table 1). Almost all interviews among the 91 selected participants took place during winter with a more even spread over the year for those not selected. Finally, participants selected for the validation study recorded a slightly lower energy intake adjusted for age and sex (p = 0.13). The table also shows some minor differences between the first and second interview of the 91 participants in the validation study in particular regarding the season of interview. Finally, the 65 participants who filled out a food diary were interviewed the second time in March 2022 which also marked the start of food diary recording, but the food diaries were filled in over several months with a median date of August 2022.

At the population level, two FFQ surveys conducted on average (median) 41 months apart (range 38–50 months) showed small and insignificant differences for energy, macronutrients and lipid fractions, whereas the calculated intake of mercury was higher in FFQ Interview 2 (p = 0.04) (Table 2). Analyses separately for men and women, three age-groups and the three locations showed similar patterns. As an example, Figure 2 illustrates the good agreement between the two surveys for total energy. As an overall measure of the completeness of the FFQ, we compared the recorded energy intake with the estimated energy expenditure and found that on average 86% of the energy expenditure was accounted for by the FFQ. For the 65 participants with a food diary, the percentage was 81%.

In Table 3 the frequency of intake of food categories is compared between FFQ Interview 2 and the food diary. The food diaries recorded 38% less meals per month than the FFQ. In the food diaries, a significantly lower frequency of intake (p < 0.001) was recorded for fruit and vegetables, dairy products, bread/pasta/rice, beverages and sugar in coffee or tea. The pattern was similar for men and women except for beverages which were more often relatively underreported in the diary among women (not shown in table). The pattern was likewise by and large similar in the three communities although both fruit and vegetables, dairy products and beverages were relatively more underreported in the food diary of participants from Nuuk (not shown in table).

In supplement table S1, the frequencies are presented for individual food items. Significant relative overreporting in the food diaries (p ≤ 0.001) was observed for certain fish (“other fish”) and pork, whereas significant underreporting (p ≤ 0.001) was observed for all categories of fruit and vegetables, cheese, rye bread, soda pop and coffee/tea with sugar.

Most food items recorded in the food diaries (88.8%) could be classified in the 45 categories of the FFQ. In addition, 10 novel food items were recorded (Table 4) but apart from “other vegetables” and eggs in rather small frequencies.

Table 4.

Food items recorded in the food diary but not in the FFQ. N = 65.

Food item	N records	Percent of all recorded items
Other vegetables	182	3.9%
Eggs	113	2.4%
Marmalade	50	1.1%
Sugar added to food	41	0.9%
Dried and canned fruit	40	0.9%
Sauce	37	0.8%
Nuts and seeds	27	0.6%
Other marine mammals	21	0.4%
Icecream	12	0.3%
Fat	4	0.1%
Total	527	11.2%

Mercury is present in high concentrations in marine food and may be used as a biomarker for a marine diet [6,7,18]. Spearman’s correlations were higher for men than for women (Table 5). For men, the highest correlation with mercury was for the frequency of consumption of marine mammals and energy percent of marine mammals (ρ = 0.50) followed by calculated intake of mercury. For women, the highest correlation was with the frequency of consumption of marine mammals.

The correlations between FFQ and blood mercury differed among geographical locations and decreased with urbanisation from ρ = 0.47/0.63 (men/women) in the capital, which has 17,000 inhabitants, to ρ = 0.26/0.27 in villages with less than 500 inhabitants (Table 6). There was an increasing gradient of blood mercury from the capital (8.8 µg/l) to the villages (38.1 µg/l).

The administration of a full FFQ is time-consuming and in some situations not necessary. Researchers have therefore sometimes limited their question to asking about the frequency of consumption of marine mammals followed by a triage of participants into low, medium and high consumers [19]. However, Table 7 shows that the agreement (Kappa) between blood mercury and tertiles of the sum of marine mammals in FFQ Interview 2 was only 0.25 (p = 0.006), but better than the agreement between blood mercury and tertiles of marine mammals in the food diary (p = 0.13).

Table 7.

Agreement between tertiles of reported consumption of marine mammals and measured mercury in whole blood. N = 60.

	Hg		Diary
	Kappa	p	Kappa	p
FFQ 2	0.25	0.006	0.20	0.03
Diary	0.13	0.17	-	-

Discussion

The FFQ was evaluated by three approaches which are discussed in turn, namely reproducibility and consistency over time, comparison with results from a food diary and the association of locally harvested marine food with blood mercury as a biomarker.

Reproducibility and consistency over time

The two FFQ surveys more than 3 years apart showed remarkably similar energy estimates and estimates of macronutrients. As illustrated with a Bland-Altman plot for energy, the individual differences between the first and the second interview clustered around zero and few were outside the 95% confidence limits. This indicates that the reproducibility of the FFQ instrument is high. The only exception was for the calculated intake of mercury which was higher in the second FFQ. This was due to a higher intake of whale meat and muktuk (whale skin), possibly caused by a difference in the season of interviews or because it has become easier to ship frozen food internally in Greenland or simply because there happened to be a larger catch of whales around the time of the second FFQ. Consumption of seal meat was similar in the two FFQs. Accordingly, at the population level the reproducibility of the FFQ was good.

Comparisons of FFQ and a food diary

FFQ and food diaries are two fundamentally different methods to estimate the intake of food. In our case, the FFQ was based on an interview and required the participants to think back and recall their usual diet, whereas the food diary was filled in by the participants on a daily basis. Overall, 38% more meals were recorded in the FFQ than in the food diary. This could reflect an overreporting in the FFQ or an underreporting in the food diary. Among the 65 participants with a food diary, the FFQ accounted for 81% of the estimated energy expenditure which is fairly standard for this method [1]. It is therefore most likely that it is a case of underreporting in the food diary rather than overreporting in the FFQ.

The agreement of the two methods was high for locally harvested food, imported meat (except pork which was underreported in the FFQ), cakes, sweets and junk food, whereas especially fruit, vegetables, dairy products, beverages and sugar added to coffee or tea was reported much more often in the FFQ than in the food diary. In some dietary studies, participants have overreported consumption of food that was socially approved, such as for instance fruit and vegetables, and total energy intake [20–22]. Overreporting due to social desirability would in principle be a potential source of bias for both FFQ and the food diary but may be more pronounced for the FFQ that was filled in by an interviewer, whereas the food diaries were filled in by the participants in private. If we assume that social desirability was most pronounced in the FFQ then the observed pattern is consistent with social desirability of eating fruit and vegetables among the Inuit similarly to the studies cited above. But it is known that locally harvested food is much appreciated by the Inuit and that eating locally harvested food is socially desirable and important for cultural identity [23–26]) and yet these foods were not overreported in the FFQ relative to the food diary. Furthermore, there is no indication in the literature that the other food groups that were reported more often in the FFQ, i.e. dairy products, beverages and sugar are socially desirable in Greenland. Other explanations must therefore exist for the relative overreporting of these foods in the FFQ. Among the most underreported food items in the diary were milk, beverages and sugar added to coffee or tea, i.e. beverages that may be underreported for technical reasons related to the way diet was reported by the two methods. Also, the FFQ interview systematically lists the food items and the interviewer prompts the participant to ponder, whereas participants who fill in a food diary may tend to forget the number of times they consume very commonly consumed food. Pork was equally underreported in the FFQ for men and women and in all three communities. It might be a case of negative social desirability but this has not been reported before.

Participants were invited to record all their food in the food diary and were thus not restricted to a limited list of items as in the FFQ. It was therefore expected that some food items appeared in the diaries that were not among the 45 food items in the FFQ. These additional food items were, however, few and not frequently reported. Only “other vegetables” and eggs were reported in more that 2% of the totally reported food items; these could be considered for inclusion in a future revision of the FFQ.

Mercury as a biomarker for consumption of locally harvested food

Mercury is found in high concentrations in meat and organs of marine mammals and is therefore a likely candidate as a biomarker for these food items and as a surrogate biomarker for locally harvested food in general. The mercury contents of food, however, vary among species and tissues and according to the age of the animal, the season and geographical location in addition to the metabolism of the individuals [1,18]. The association between whole blood mercury and diet is therefore not straightforward. We observed a statistically significant correlation between the results of the FFQ and measured mercury in whole blood but the Spearman’s rho was not high; 0.50 for the highest correlation for men and 0.41 for women compared with previous findings (Pearson correlations) from Greenland of 0.57 for men and 0.29 for women [6] and 0.56 for men and women [7]. The correlation between FFQ and measured mercury differed by FFQ measure, i.e. frequency of meals, energy percent and calculated intake of mercury and by food items such as seal meat, sum of marine mammals, fish and sum of all kalaalimernit. The highest correlations were found for the sum of marine mammals and the lowest for fish but the correlations varied little among frequency, energy percent and calculated mercury intake. The inclusion of portion sizes and calculations of energy and mercury thus added little to the explanatory power of the FFQ.

There are several explanations for the low correlation between FFQ and blood mercury. The FFQ has only questions about consumption of seal meat, whale meat, muktuk and fish while the highest concentrations of mercury are found in liver and kidney of marine mammals and marine birds [11]. The FFQ asked about the usual diet while mercury has a half life in the body of 50 days [27] so the blood mercury only reflects the mercury burden for a short period of time. Since there is pronounced seasonal variation in the availability of locally harvested marine food, this may add to the discrepancy between FFQ and measured mercury. The inverse correlation of blood mercury and Spearman’s correlation at the community level is worth studying further.

Reduced to comparisons of tertiles, the Kappa statistics for agreement between recorded intake of marine mammals and measured blood mercury was not impressive. It was higher for FFQ (κ = 0.25) than for the food diary (κ = 0.13) which suggests that FFQ was a better proxy for consumption of marine mammals than was the food diary.

A number of population health studies have been carried out recently in the Circumpolar North, for instance among the Inuit in Nunavik, Canada, and among the Sami in northern Norway [28,29]. The researchers have benefitted from close cooperation regarding dietary and other studies.

Conclusion

At the population level, the 45-item FFQ used in Greenland is reproducible to a satisfactory degree and judged by blood mercury as a biomarker for marine food, it reflects the distribution on local and imported food fairly well. Compared with a food diary there were a number of discrepancies regarding the frequency of consumption of food groups and some additional food items were identified in the food diary. The results are important in the process of revising the FFQ, for example, by adding new food items identified in the food diary. The FFQ was designed for use among the Greenland Inuit who have a specific pattern of locally harvested and imported food and it is not expected to function as well in other cultural settings not even among the Inuit in Canada and Alaska who rely on other locally harvested species and different types of imported food than in Greenland. But the study is important as an inspiration for other countries and ethnic groups who may benefit from the methods used to create, validate and revise the Greenland FFQ.

Funding Statement

The work was supported by the Greenland Research Council.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Ethics

Ethical approval of the study was obtained from Greenlandic Ethics Committee (KVUG 2017–05 and 2020–20) and written informed consent was obtained from all participants.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/22423982.2024.2332008.