Food sources of energy and nutrients among Canadian adults following a gluten-free diet

Background The gluten-free diet (GFD) involves the elimination of wheat and related grains. Wheat is a key fortification vehicle for nutrients such as iron and B vitamins. While there is growing evidence of low nutrients intake and poor diet quality amongst people following long-term GFD, few studies have used a dietary pattern approach to analyse top food sources of nutrients in today’s complex food environment. Thus, the purpose of this study was to identify food sources of energy and nutrients from previously collected diet records of adults following a GFD. Methods Three, 3-day food records were collected from 35 participants in a lifestyle intervention study (n = 240 records). All food items were categorised according to the Bureau of Nutritional Sciences Food Group Codes. Percentages of total dietary intakes from food groups were ranked. Results Mean intakes of dietary fibre, calcium and iron (females) were lower than recommended, with half the sample consuming below the recommended proportion of energy as carbohydrate. Meat, poultry and fish were the top source of energy (19.5%) in the diet. Gluten-free (GF) grain products were the top source of carbohydrate, fibre and iron and second greatest source of energy. Amongst grains, breakfast/hot cereals, yeast breads, and mixed grain dishes were the greatest nutrient contributors, despite most commercial cereals and breads (65%) being unenriched. Legumes were not frequently consumed. Conclusions GF grains were the top food source of carbohydrate, fibre and iron, despite few brands being enriched or fortified. It is a challenge to assess and monitor nutrient intakes on GFD due to the lack of nutrient composition data for B vitamins and minerals (other than iron). Dietary planning guidance for the appropriate replacement of nutrients provided by wheat is warranted.


INTRODUCTION
Celiac disease (CD) is a systemic, autoimmune disease triggered by the ingestion of gluten in genetically susceptible individuals (Leonard et al., 2017). The inflammatory process targets the intestine, leading to progressive villi damage and, consequently, malabsorption of nutrients, particularly iron, folate, calcium and fat-soluble vitamins (Theethira, Dennis & Leffler, 2014). Multiple micronutrient deficiencies are common at diagnosis and identify food sources of energy and nutrients of concern. Findings will inform patterns of dietary adequacy for people requiring a GFD.

Population selection
Details on participant recruitment and the intervention protocol were previously reported (Dowd et al., 2019). Briefly, a convenience sample of volunteers was recruited through physician referral, online and local advertising. Inclusion criteria were: a self-reported CD diagnosis (through blood test and/or biopsy) or gluten intolerance, physical inactivity, residence in the Calgary (Alberta, Canada) area, and an age of at least 18 years. Thus, participants were on a GFD prior to entering the study. The primary outcomes were quality of life and gut microbiota composition following 12 weeks of exercise training. Diet records were used for secondary analyses. The full study protocol was approved by the University of Calgary Research Ethics Boards (REB 16-17740-MOD4). Informed, voluntary, written consent was obtained prior to participant enrolment. Approval for secondary analysis of data was granted by the St. Francis Xavier University Research Ethics Board (Romeo #23766).

Data collection
Demographic data were collected through an online questionnaire. Written, self-reported dietary data from 35 participants was collected using 3-day weighed food records on three separate occasions, resulting in a total of eighty 3-day records, or 240 days of food intake. Dietary data was collected before (n = 35) and immediately after the 12-week intervention (n = 23), with a follow-up assessment 3 months later (n = 22). A Registered Dietitian entered all dietary data into the nutrient analysis software FoodWorks (Version 18.0; The Nutrition Company, Long Valley, NJ, USA), linked to the 2015 Canadian Nutrient File (CNF). Missing nutrient values were imputed using the CNF, a brand-specific nutrition fact panel (NFP) or assumed to be zero based on the ingredient list, when appropriate. There were no missing values in the final nutrient data file. Supplements recorded in the food records were excluded from further analysis.

Data analysis
Willet's criteria were applied to evaluate plausible energy reporting, and no food records were excluded (Willett, 2012). Descriptive statistics were carried out on nutrients intake. Mean nutrients intake was compared to the Dietary Reference Intakes to provide context in interpreting the food sources of nutrients and whole grain consumption patterns in this sample population. Mean nutrients intake below the Estimated Average Requirement (EAR), recommended daily allowance (RDA), or Adequate Intake, or above the Tolerable Upper Intake Level, were identified as nutrients of concern. After combining all time points, the final food item list contained 4,189 discrete entries. Each food was classified using the Bureau of Nutritional Sciences (BNS) Food Group Codes and Descriptions developed for the Canadian Community Health Survey. Major categories of foods were adapted from the USDA Dietary Source Nutrient Database, as previously reported (Jamieson, Rosta & Gougeon, in press), to reduce 78 BNS groups and sub-groups into eleven main categories for aggregate analyses (Table S1). Food items were classified as reported, rather than disaggregated, as recipe and ingredient-specific data were not available. Combined food items (e.g. pizza, sandwich) were captured under the BNS 'Mixed dishes' sub-categories for grains, meat, poultry and fish (MPF), dairy, vegetables, or fruit. Minor adaptations to the BNS categories were made, as previously described (Jamieson, Rosta & Gougeon, in press), to ensure comparability to a previously described Canadian GF population. For each food grouping and sub-category, cumulative nutrient values were determined and expressed as a percentage of total nutrient amount. Food groupings were ranked in order from highest to lowest according to nutrient contribution percentage. All statistical analyses were conducted in StataSE 14.2 (StataCorp, 2015. College Station, TX, USA).

Participant characteristics
Participants with valid dietary data included 35 adults (83% female) with CD, residing in Calgary, Alberta. Mean age was 47 ± 11.5 years, body mass index was 28.2 ± 5.2 at the start of the study and average time since diagnosis was 6.7 ± 6.0 years.

Nutrients intake
No difference in mean energy intake was observed across the three time points so all dietary data was combined for further analyses (Table 1). Macronutrients intake were, on average, 45.2% from carbohydrate, 37.2% from fat and 17.5% from protein. Mean intakes of calcium and vitamin C did not meet the EAR. Mean intakes of iron were between the EAR and RDA for females <51 years and mean dietary fibre intakes fell below the AI. All other mean nutrients intake exceeded the RDA, while mean sodium intake surpassed the UL.

Food sources of nutrients
Food group sources of energy, protein, iron, carbohydrate, dietary fibre and total sugar are shown in Tables 2 and 3. Food sources of fat, calcium, sodium, vitamin C, vitamin A, cholesterol and saturated fat are available online as Tables S2-S6.
Top energy sources included MPF (19.5%), GF grain products (19.0%), dairy products (12.5%), desserts and sweets (8.9%) and beverages (8.0%) ( Table 2). Although MPF was the highest ranked food group for energy, two of the top sub-categories including mixed dishes, which also incorporated GF breads (e.g. hamburgers), potatoes and/or vegetables, grains (e.g. rice or noodle stir fry). Poultry was the highest ranked non-mixed MFP sub-category for energy. For grain products, yeast breads, breakfast and hot cereals, mixed grain dishes (pasta and rice-based), and rice were the largest energy contributors, in descending order. Legumes, nuts and seeds (LNS) contributed 5.5% of energy.
Top sources of protein were MPF (45.1%), dairy products (17.9%), followed by grain products (10.1%) ( Table 2). The greatest food sources of iron were grain products (22.8%) (primarily cereals), MPF (22.6%), vegetables (14.6%) and LNS (7.7%) in descending order (Table 2). Breakfast and hot cereals included oatmeal or oats (35.6 % of items) and seven commercial ready-to-eat cereal brands (three of which were fortified with vitamins and minerals). Yeast breads included 13 different commercial brands and nine generic CNF items. Four of the 13 commercial bread brands were enriched. Mixed grain dishes included 10 fried rice dishes, seven rice and cheese dishes, five pasta-based dishes (with cheese, vegetables or meats). As a percentage of all items, the LNS category included 33.9% of items as seeds (e.g. psyllium, hemp, chia, flax and pumpkin), 27.0% nut items (e.g. almond, cashew, walnut), 23.8% nut butters (mainly from peanut), 6.4% chickpea or hummus and 3.7% beans.
In descending order, the top sources of carbohydrates were grain products (30.0%), fruit (11.9%), vegetables (11.8%) and desserts and sweets (11.1%) ( Table 3). Breakfast and hot cereals, yeast breads, mixed grain dishes, and rice were the greatest carbohydrate contributors among grain products. LNS contributed 3.0% of carbohydrate intake. The top sources of dietary fibre were grain products (23.2%) (from breakfast and hot cereals, yeast breads and mixed grain dishes), vegetables (22.0%) and fruit (16.8%) ( Table 3). LNS ranked fourth in fibre contribution (13.3%), mainly in the form of nuts and nut butters.  The top ranked sources of sugar were fruit (20.6%), desserts and sweets (18.6%) and beverages (17.4%) ( Table 3). Grain products ranked fifth as a source of sugar (9.7%).

DISCUSSION
The GF food environment is complex and rapidly changing, with new and diverse food options for emerging populations of consumers and patients seeking GF foods. While the poor diet quality of packaged GF foods has been previously established (Elliott, 2018;Jamieson, Weir & Gougeon, 2018;Kulai & Rashid, 2014), actual food sources of nutrients in the GFD have not been thoroughly explored. Thus, this study presents novel findings on the top nutrient sources for a sample of Canadian adults diagnosed with CD or gluten intolerance. In agreement with GFD assessment studies (Vici et al., 2016), lower mean intakes of dietary fibre, calcium and iron (for women) from food sources were observed in this study, but most mean nutrient intakes met recommendations. Macronutrient balance was shifted from the recommended ranges towards lower energy intake from carbohydrate (with half of participants consuming less than 45%) and higher energy intake from fat (with over half of participants consuming more than 35%). This is exemplified by lower mean intake of dietary fibre and the higher ranking of energy from MPF over grains. The mean 45% energy from carbohydrate is lower than previously observed in Nova Scotia (NS)  adults consuming a GFD (50%), despite a similar sex balance in study samples (Jamieson, Rosta & Gougeon, in press). Similarly, a study of diet-experienced people with CD (n = 55) in Australia reported 45.8 ± 0.7% energy as carbohydrate (Shepherd & Gibson, 2013), whereas a British diet study (n = 49) reported a 48% average (Wild et al., 2010). Thus, average carbohydrate intake in the present study is at the lower end of observed intakes in CD. This may partly be related to the regional dietary pattern which trends toward lower relative carbohydrate intakes than the national average (Statistics Canada, 2019).
Additionally, this could reflect other participant characteristics such as lack of food skills or literacy with respect to wheat replacement or lack of support and dietary guidance for those requiring GFD, as previously noted (Jamieson & Gougeon, 2019). Approximately two-thirds (67.9%) of the energy consumed by this sample was derived from five food groups (MPF, grain products, dairy products, desserts and sweets and beverages). Notably, this resulted in a higher ranking of MPF and lower ranking of fruits and vegetables in this sample compared to the NS GF study (Jamieson, Rosta & Gougeon, in press) and was reflected in the nutrients with lower mean intakes (fibre, iron). The high-ranking contribution of dairy foods (third) to energy intake is not reflected in the mean calcium intake observed. However, this may be explained by the emphasis on cheese and eggs (4.9% and 3.0% of energy, respectively) and lesser emphasis on yogurt and milk. LNS ranked quite low (eighth) on the contribution to energy intake, in agreement with the NS GF Study (seventh) (Jamieson, Rosta & Gougeon, in press).
Of all grain products, cereals were the top contributor of carbohydrate, dietary fibre and iron in agreement with the NS GF Study (Jamieson, Rosta & Gougeon, in press). At the same time, grain products were fifth in contribution to total sugar intake, with cereals as the top source. Cereal as a top source of fibre is likely related to frequent consumption of oats and unenriched, ready-to-eat cereal brands which tended to be made from whole grains and 'natural' ingredients. In contrast, enriched ready-to-eat cereals were also frequently consumed, likely contributing to micronutrient intake, but not fibre. Thus, GF cereals appear to represent key foods with two distinct nutrient profiles, varying with enrichment practices and fibre-rich ingredients. Notably, vegetables, fruits and nuts were greater dietary fibre contributors than cereals. Gluten-free breads were also a key contributor to carbohydrate and dietary fibre, and to a lesser extent, iron intakes, despite being primarily unenriched (89.3%). This may be related to the growing food trend to incorporate ancient grains and seeds in the North American market (Roesler, 2018). Thus, while GF breads often use rice or starch as the first ingredient, other whole grain flours (e.g. amaranth, quinoa) and seeds (e.g. chia, flax) may be added as secondary ingredients, improving nutrient density. However, while ranking highly amongst grain products, the absolute contribution of iron was still relatively small for breads.
A mixed grain dish category (rice or pasta-based) was the other top grain sub-category source of carbohydrate, dietary fibre and iron. However, nutrient contributions from this category are difficult to interpret since additional ingredients (e.g. meats, vegetables) are also included. In Canada, the enrichment of rice and non-wheat pasta is voluntary (Government of Canada, 2019a) and thus not a reliable micronutrient source. The selection of brown rice, brown rice noodles, and whole grain-based GF pastas could enhance fibre and select micronutrient (e.g. magnesium, zinc) intakes (Government of Canada, 2015). In addition, non-grain key contributors of iron (e.g. LNS and select vegetables) could be emphasised to improve intakes for those following the GFD. In fact, these suggestions align well with Canada's Food Guide, which emphasises a central role for vegetables and fruits, along with whole grains and plant-based protein foods such as beans, lentils, nuts and seeds (Government of Canada, 2019b). In the present sample, LNS made a small contribution (<8%) to energy, carbohydrate, iron and protein, but a moderate contribution (13.3%) to dietary fibre. Most LNS choices in the present study were from various seeds, nuts and seed/nut butters, contributing more fat than fibre. In the present study, and in the NS GF Study (Jamieson, Rosta & Gougeon, in press), infrequent legume consumption was observed, an especially rich plant source of naturally occurring vitamins and iron (Government of Canada, 2015).
This study has several important limitations. Due to the nature of convenience samples, the dietary patterns observed cannot be generalised. However, the large number of diet records collected over several time points helps to counter this problem. The lack of ingredient-specific data was a limiting factor in food categorisation. The lack of brand specific GF nutrient composition data (e.g. B vitamins, minerals) limited the number of nutrients that could be reliably analysed. A control group was not available for comparison purposes, however some national data on food sources of nutrients are available for the general Canadian population (Garrigeut, 2011;Langlois & Garrigeut, 2011;Statistics Canada, 2017). As with any food coding system, results are affected by categorisation choices and thus, comparisons between studies should be done with caution. However, this study did follow the same protocol as the NS GF Study and is comparable in that respect. Finally, there were some missing diet records (19%) which may have influenced the types of foods reported in this sample.

CONCLUSIONS
Although carbohydrate accounted for only 45% of energy intake, on average, in this sample of adults following a GFD, grain products were still a top source of nutrients including carbohydrate, dietary fibre and iron. Top grain contributors included cereals and yeast breads, despite few of the brands reported being fortified or enriched with micronutrients. Thus, while GF grains are a source of fibre, they are less reliable as key sources of vitamins and minerals. Given this and the lack of nutrient composition data for micronutrients (other than iron) in GF foods, it would be prudent to re-evaluate enrichment and fortification policies, which presently are optional for GF flours in Canada. In addition, improvements in the nutritional quality of GF food formulations could contribute to nutrient adequacy for those requiring long-term GFD. Findings also suggest that legumes, a good source of folate, potassium, minerals, fibre and protein, were not a key nutrient source in this study. As people with CD are at increased risk of vitamin and mineral deficiency (Rondanelli et al., 2019), nutritional adequacy of the GFD will require dietary planning and appropriate replacement of nutrients provided by enriched wheat products. Without reliable exposures to key micronutrients through population-level fortification, nutritional monitoring and dietary guidance from a registered dietitian appears even more critical. Further research is needed on nutrient composition data in novel GF foods and ingredients, as well as in regard to the accessibility of registered dietitians to support people requiring long-term GFD management.

Human Ethics
The following information was supplied relating to ethical approvals (i.e. approving body and any reference numbers): The University of Calgary Research Ethics Board (REB 16-17740-MOD4) and St. Francis Xavier University Research Ethics Board approved this research (Romeo #23766).

Data Availability
The following information was supplied regarding data availability: The raw nutrient data for each food item reported is available in File S1.

Supplemental Information
Supplemental information for this article can be found online at http://dx.doi.org/10.7717/ peerj.9590#supplemental-information.