Aliment Pharmacol Ther 2010; 32: 573–581
Background The only therapy for coeliac disease (CD) is a long-term gluten-free diet (GFD). Little is known about the detailed composition of such a diet.
Aim To clarify the nutritional composition of a GFD and to compare it with a non-GFD diet in representative non-CD populations.
Methods A total of 139 consecutive patients with CD were invited to fill in a prospective validated 5-day food diary, of whom data from 93 were analysed. Results were compared with data from the National Diet and Nutrition Survey of Adults and the UK Women’s Cohort Study (UKWCS).
Results Individuals consuming a strict GFD generally had similar intakes of energy and nutrients to those of comparison populations, but a higher proportion of carbohydrate intake was obtained from nonmilk extrinsic sugars and intakes of nonstarch polysaccharides were low. Compared with the UKWCS sample, female patients adhering to a GFD had lower intakes of magnesium, iron, zinc, manganese, selenium and folate. In male patients, intakes of magnesium and selenium were particularly low.
Conclusions This study reinforces the need for clinicians to recognize that avoidance of gluten cannot be the sole focus of a gluten-free diet. Maintenance of adequate intakes of essential nutrients and in particular complex carbohydrates must also be the goal for patients.
Increasing numbers of people are being diagnosed with Coeliac Disease (CD),1 for which the only available treatment is a lifelong gluten-free diet (GFD).2 As the GFD has to be maintained for life, any imbalances in the diet would have significant implications for the overall health of a ‘treated’ patient with CD. Although there have been numerous studies of the benefits to patients of going from a gluten containing diet to a GFD, on being diagnosed with CD,1 most studies considering the nutritional adequacy of the gluten-free diet have been based outside the UK3–7 and there are little data on the detailed composition of GFDs in the UK. Kinsey et al.8 reported on intakes of 49 patients with CD in the UK. They found inadequate intakes of nonstarch polysaccharide (NSP), calcium and vitamin D.
In a recent systematic review of previous research in the area, Robins et al.9 highlighted that when considering iron, calcium and B vitamins, there is no evidence base to suggest that nutritional deficiency is a significant problem in individuals diagnosed with CD. However, neither is there any high level evidence that individuals with CD have good nutritional status. Furthermore, all eleven studies considered as part of the systematic review were considered to be at moderate or high risk of bias.9
A GFD should meet individual requirements and Dietary Reference Nutrient Intakes (RNIs) for the general population.10 There may be a case for increased nutritional requirements in those with CD,9 but there are currently no specific nutritional recommendations other than for calcium.11
The aim of this study was to investigate and quantify the nutritional content of the GFD in order that its nutritional adequacy can be assessed.
The objectives to realize this end are:
To obtain high quality dietary information from a well-characterized group of treated coeliac patients, on a strict GFD, using a validated 5-day food diary.
To identify how and what proportion of patients on a GFD are meeting the UK recommended intakes for a number of macro and micronutrients by comparing the dietary analysis with RNIs10 and with specific recommendations for calcium in those with CD.11
To compare the nutritional content of a GFD in the UK with a normal (gluten containing) diet, by comparing with National Diet and Nutrition Survey (NDNS)12 data for the general (Northern) population and an age/gender-matched cohort from the United Kingdom Women’s Cohort Study (UKWCS).13
While a weighed food intake is considered by some to be the gold standard for assessing dietary intake,14 it has a number of disadvantages including the time it takes to complete and the inaccuracy for people consuming foods outside the home or those not prepared in the home. Nutritional studies in the UK arms of the European Prospective Investigation of Cancer (EPIC) study use a validated (against both 16 day weighed records and urinary and serological biomarkers) portion size-based food diary to assess dietary intake that minimizes some of these limitations.15
Ethics approval was obtained from Leeds (East) Research Ethics Committee.
Participants all had histologically confirmed CD (i.e. at least Marsh I criteria 16) and had been on a strict GFD for 6 months or more prior to being invited to take part in the study. Dietary compliance was assessed by self-reporting and dietician review. Available coeliac serology was also assessed, although negative serology was not an absolute prerequisite for consideration of entry into the study. Previous studies suggest between 7% and 10% of coeliac patients self-reporting high compliance with a GFD still have positive serology.2
Patients who met the inclusion criteria were recruited by a dietician in a weekly specialist gastroenterology clinic, or specialist coeliac clinics, in a large UK teaching hospital over an 18 months period from January 2007 to May 2008. Written consent was obtained. Participation was voluntary. Participants were invited to complete a detailed 5-day food diary, to include 3 weekdays and 2 days at the weekend. The EPIC validated food diary was used,15 which includes colour photographs for estimating food portion sizes. The dietician gave detailed completion instructions at the time of recruitment and a follow-up telephone call was made in the week following recruitment. Diaries were returned by post and telephone contact was made to clarify details, where necessary.
Participants’ weights and heights were recorded in clinic at recruitment for use in calculating energy requirements and body mass index (BMI). All patients also had routine bloods and samples for micronutrient levels taken (data not shown).
The diaries were analysed using Microdiet version 2.52 (2005; Downlee Systems Ltd, Chapel le firth, UK), a computerized nutrient data bank, which has details of Glutafin ‘Gluten Free Dietary’ range built-in. Data for gluten-free foods provided by other manufacturers were also added to the database.
Portions sizes were calculated from the photographs in the EPIC food dairies using conversion data from the EPIC group,17 with weights provided by participants or calculated from household measures using data from the FSA Food Portion Sizes Guide.18 Nutrients provided by supplements were not included in the nutritional analysis, although details of dietary supplements consumed were recorded and reported separately.
A comparison was made between reported energy intakes (calculated from the 5 day food diary) and estimated individual energy requirements based on individual activity levels and basal metabolic rate (BMR – which in turn is estimated using Schofield equations). As the original design of the study had not included asking participants about their activity levels, an assumption was made that all patients were sedentary, which equates to a physical activity level (PAL) of 1.4. This calculated estimated individual energy requirement was used to determine the level of under- and over-reporting in the food diaries. A threshold of ±20% of estimated individual energy requirement was used to assess level of under- and over-reporting in the returned food diaries; no adjustment was made for BMI. The lack of data regarding activity levels had no direct bearing on the interpretation of the data contained within the food diaries – only on the assessment of under- and over-reporting.
Results obtained were compared against RNI10 and in the case of calcium, the British Society of Gastroenterology (BSG) guidelines for CD.11 Nutrient intakes were also compared against a comparable population from both the NDNS: Northern Region and the UKWCS.
Nutrient data from 195 male and 256 female participants of the National Diet and Nutrition Survey selected from Northern region were sampled to provide a noncoeliac reference population. The NDNS which was carried out between June 2000 and July 2001 presents the most recent dietary intake data for a UK population between the ages of 19 and 64 years.12 In addition, it provides regional data and where available, these data were used as being most comparable with the study population.
The UK Women’s Cohort Study (UKWCS) was started in 1993 to investigate diet and cancer relationships in middle-aged women in the UK. A food frequency questionnaire and 4 day semi-weighed food diary were used to assess dietary intake.13 Participants had been initially selected to ensure a wide range of dietary patterns were represented (including approximately one-third following a vegetarian diet) and at time of the completion of the original questionnaire were aged 35–69. A sample of 708 randomly selected cancer-free cohort participants who had a fully coded food diary were available as a healthy noncoeliac reference population. Nutrient intakes were estimated from food diaries using in-house MS Access-based software based on the Fifth Edition of the Composition of Foods and related supplementary food tables.19
A student’s unpaired t-test was used for comparison of means: GraphPad software (GraphPad Software Inc., La Jolla, CA, USA) was used, with a P-value <0.05 deemed as statistically significant.
A total of 139 patients were recruited into the study; 100 (72%) food diaries were returned (mean age 54 years; median age 55 years). Of the 39 diaries not returned (23 women, 16 men), mean age was 41 years and median age was 43 years. Seven of the returned diaries were excluded as they did not meet the inclusion criteria on return: three diaries included foods containing significant amounts of gluten (one man, two women) and four contained insufficient detail and could not be analysed accurately (three male, one female). Therefore, 93 diaries were included in the study; 62 (67%) by female patients. Female patients comprise 69% of coeliac clinic attendees.
Mean age for female patients was 53 (21–79) years and for male patients 56 (18–74). Mean BMI was 25.5 and 25.4 respectively. For more detailed data, see Table 1.
|Mean BMI (s.d.)||25.4 (4.0) n = 25||25.5 (3.9) n = 45|
|BMI range (kg/m2)*||17.8–33.6||18.4–35.1|
|Mean age (years)||56 (15) n = 31||53 (13) n = 62|
|Median age (years)||63||53|
|Mean duration of GFD (years)||8.43||11.28|
|Median duration of GFD (range)||6.71 (0.29–31.35)||9.52 (0.52–33.45)|
Of the 93 patients, 54 had coeliac serology (anti-tTG) from when they submitted their food diaries. Of these, only six (11%) were positive (two were described as ‘weak positive’ and three of the remaining four positives also had positive anti-endomysial antibodies). No patient with positive anti-tTG antibodies reported eating gluten in their diary and therefore this group was not analysed separately.
Patients were asked about other conditions or dietary practices influencing their dietary intake (in addition to gluten restriction). One male patient reported trying to gain weight during the recorded week and four men had other health conditions including type 2 Diabetes Mellitus (DM), pancreatic cancer, and enteropathy-associated T-cell lymphoma. Fifteen of the women reported other wide-ranging conditions (including type 1 DM and fibromyalgia, irritable bowel syndrome, Sjogrens syndrome, multiple sclerosis, Parkinson’s disease, idiopathic chronic constipation, thyroid disease and ulcerative colitis) that could have affected their dietary intake. One female participant was recorded as vegetarian. Five other participants excluded other dietary components such as lactose, soya or eggs and two reported trying to lose weight during the week recording their diaries. The 15 women with reported co-morbidities had a mean energy intake of 7486 kJ (s.d. 1495) compared to 8018 kJ (s.d. 1872) for those with no reported co-morbidities (P = 0.32). For the four men with reported co-morbidities mean energy intake was 9244 kJ (s.d. 3329) compared to 10572 kJ (s.d. 2977) for those with no reported co-morbidities (P = 0.42) i.e. there were no statistically significant differences in overall energy intake when the data were stratified according to presence or otherwise of co-morbidities. Therefore, all subjects were included in the data analysis regardless of their health status. Co-morbidities were not recorded in NDNS and only minimally in UKWCS.
Determining the level of under- and over-reporting of energy intake. A threshold of ±20% of estimated individual energy requirement was used to assess level of under- and over- reporting; no adjustment was made for BMI. Based on this assumption, 26% men and 23% women under-reported and 23% men and 13% women over-reported their intake. No food diaries were omitted on the basis of this analysis.
Macronutrients. Detailed data are given in Tables 2 and 3, with numbers of patients meeting overall recommendations being given in Table 4. In summary, women on a GFD consumed significantly (P < 0.05) more energy across all macronutrients (i.e. protein, fat and carbohydrate) than an age- and gender-specific ‘local’ population i.e. Northern cohort from NDNS. Compared with UKWCS participants, study women had significantly (P < 0.05) lower intakes of fibre (NSP), but similar intakes of macronutrients. As only one of our female participants declared herself as vegetarian (in contrast to 27% of those in UKWCS), we compared NSP intakes in those women on a GFD with only nonvegetarians in UKWCS. There was still a significantly lower mean intake of NSP in women in our study (13.7 g/day and 16.5 g/day respectively; P < 0.001). Male patients on a GFD also consumed significantly (P < 0.05) more energy (fat and carbohydrate, but not protein) compared with the local NDNS population, but had a similar (low) intake of dietary fibre (NSP).
|Units||GFD||NDNS Northern (n = 256)||UKWCS (n = 708)|
|Mean||s.d.||Mean||s.d.||P value (vs. GFD)||Mean||s.d.||P value (vs. GFD)|
|Starch and intrinsic sugar||g||168.8||41.5||152||NA|
|Units||GFD||NDNS Northern (n = 195)|
|Mean||s.d.||Mean||s.d.||P value (vs. GFD)|
|Starch and intrinsic sugar||g||220.9||72.1||196||NA|
|RNI men/women||n (%) meeting RNI|
|Men (n = 31)||Women on GFD (n = 62)||UKWCS (n = 708)|
|Folate||200/200 μg||28 (90)||49 (79)||610 (86)|
|Calcium||700/700 mg||23 (74)||42 (68)||500 (71)|
|Calcium (coeliac)||1000 mg ≤55 years||n = 13; 7 (54)||n = 34; 11 (32)|
|1200 mg >55 years||n = 18; 7 (39)||n = 28; 5 (18)|
|Magnesium||300/270 mg||7 (23)||19 (31)||500 (71)|
|Zinc||9.5/7 mg||17 (55)||37 (60)||498 (70)|
|Iron||8.7/8.7–14.8 mg||25 (81)||Aged 18–54 RNI = 14.8 mg: N = 31|
Aged 55–74 RNI = 8.7 mg : N = 27
|3 (10)20 (74)||606 (86)|
|Selenium||75/60 μg||2 (6)||7 (11)||247 (35)|
|Manganese||2.3/1.8 mg||15 (48)||47 (76)||678 (96)|
|Protein||55.5-53.3/45–46.5 g||Aged 18–50 RNI = 55.5 g: N = 11|
Aged 50–74 RNI = 53.3 g: N = 20
|Aged 18–50 RNI = 45 g: N = 27|
Aged 50–74 RNI = 46.5 g: N = 35
|Energy as Fat||<33%/<33%||17 (55)||33 (53)||376 (53)|
|Energy as CHO||>47%/>47%||13 (42)||26 (42)||338 (48)|
|Energy as extrinsic sugar||<10%/<10%||8 (26)||15 (24)|
Micronutrients. Female patients on a GFD consumed significantly (P < 0.05) more calcium and magnesium than an age- and gender-matched ‘local’ population i.e. Northern cohort from NDNS. Compared with UKWCS, the women had significantly lower intakes of magnesium, iron, zinc, manganese, selenium and folate (P < 0.05), although there was no overall difference in the percentage of the population meeting dietary reference values (DRVs) for nutrient intake apart from significantly lower numbers of women in this study meeting DRVs for magnesium (31% vs. 71%) and selenium (11% vs. 35%). However, for calcium intake, only 32% of under 55 s and 18% of over 55 s of female patients with CD met current guidelines.11 There was no comparable population for men with regard to micronutrients. Looking at percentage of men meeting DRVs, they also did not meet recommended intakes for magnesium and selenium (23% and 6% respectively), and the level of recommended manganese intake was also low (48%). Levels of recommended calcium intake were not as low as in women. There is no UK recommended intake for Vitamin D as a large proportion is synthesized from sunshine – indeed Millen et al. suggest that total vitamin D intake from foods and supplements explains only 7% of the variance in 25(OH)D concentrations between participants of the Women’s Health Initiative Trial.20 More details are given in Tables 2–5.
|Men ≤55 years||13||619.1||2269.7||1160.3||496.8|
|Men >55 years||18||455.0||2005.3||1040.5||437.6|
|Women ≤55 years||34||259.1||1716.8||862.2||374.4|
|Women >55 years||28||334.8||1970.8||920.1||313.4|
Dietary supplementation. Just under half of the GFD study participants reported taking a combined calcium and Vitamin D supplement (47% women and 48% men) the majority having obtained these by prescription. In addition, two women stated that they had calcium and vitamin D supplements prescribed, but did not take them. A smaller proportion of participants reported taking iron supplements [11 (18%) women and 2 (7%) men], and of these, most had received them on prescription.
There was no correlation between dietary intake and dietary supplementation i.e. those with low dietary intake were no more likely to be taking a supplement than others.
This study shows that female patients on a GFD compared with the Northern region population of the NDNS, (Table 2) consumed significantly more energy (fat, CHO and protein), more calcium and more magnesium (P < 0.05). When comparing the intakes with the UKWCS cohort, the study female patients had significantly lower intakes of fibre (NSP), magnesium, iron, zinc, manganese, selenium and folate (P < 0.05) with no significant difference in energy intake. This suggests less nutrient dense energy food choices by women in this study, such as sugary snack foods. Comparing intakes of NMES supports this. Whilst the UKWCS has no data for NMES, comparison with the NDNS shows statistically significant higher intakes of NMES intake in those on a GFD (P < 0.01). For the female patients with CD, 54% of the additional carbohydrate comes from NMES and for the men, 35.7%. The recommended intake of NMES is 10% total energy intake21 (Table 4).
In men, a comparison of macronutrient intake with that of the NDNS again reveals that men in this study consumed significantly more energy (fat and CHO but not protein), (P < 0.05) (Table 3). Notwithstanding this, only 42% of men and women got more than 47% of their total energy intake from carbohydrate sources (Table 4). It is recommended that at least 47% of total energy intake comes from carbohydrate sources.10 These data and the extrinsic sugar figures confirm that patients on a GFD are not consuming enough complex carbohydrates.
Calculation of estimated individual energy requirements indicates 22 (35%) of the reported intakes from the women and 15 (48%) of the male-reported intakes fall outside ±20% of their requirements. This is comparable to similar studies.6, 22 Bingham23 suggests an average error of ±10% for calculated energy intake and states, ‘Habitual energy intakes from dietary surveys of groups of adults 72% or less than BMR are almost certainly invalid’. None of the intakes in this study was below this threshold. Hallert et al.6 use a similar method for validating their data and suggest that patients underestimated intake by about 25% which is a little more than levels found in this study. The NDNS-reported intakes are compared with estimated average requirement (EAR), which does not take into account weight or activity, unlike the estimated individual energy requirement. This comparison suggests that the men (all regions) reported an intake on average 92% of the EAR and women, 85% EAR. Corresponding figures for this study are 102.2% and 98.3% i.e. in this study, men ‘over-report’ by only 2.2% and women ‘under-report’ by only 1.7%.
We collected data using a 5-day food diary including a weekend (or two nonworking days), whereas previous studies3–8 used three or 4 days diaries. Levine and Morgan24 suggest that a week is too short to assess intakes of vitamins, minerals and trace elements. However, Livingstone et al.25 argue that recordings lasting longer than 4 days may not necessarily be superior.
Strengths of this study are that it is comprehensive, with nutritional composition of gluten-free foods verified by contact with manufacturers. All (bar one) of the patients in this study had been on a GFD for at least 6 months – median time on a GFD was nearly 7 years for men and 9.5 years for women.
Unlike many other studies,9 this study does present data of co-morbidities and compares intake with comparable populations.
Weaknesses of this study include the relatively small sample size (although the authors are not aware of any larger cohort of patients on a GFD that have been dietetically assessed). Previous similar studies3–8 have sample sizes ranging from 30 to 71 participants. The cohort in the UKWCS by definition is a health conscious cohort of middle aged women, and a higher micronutrient intake than those on a GFD could therefore be anticipated. However, as the study participants have maintained a rigid diet (GFD) and been regular attendees of a specialist clinic, it could be contended that this group are ‘health conscious’ too. As comparator populations, younger ages were under-represented in the UKWCS cohort and in NDNS, there is concern about the probable under-reporting of nutrient intake (given the overall reported energy intakes). In NDNS too, the younger age ranges were under-represented – of the women, only nine were aged 18–40 years.
Significant differences between GFD and non-GFD groups of consumption of home-prepared and processed food would skew our data, and it is generally easier for non-GFD patients to ‘snack’ and eat processed foods spontaneously, especially outside the home. Data with regard to food preparation is incomplete, but a random subgroup of 30 diaries was reanalysed, and of these, only five (17%) patients had consumed a ready meal during the 5 days covered by the food diary. However, in percentage terms, 47% of energy intake was attributable to processed foods in those on a GFD. There are no such equivalent data in UKWCS, although NDNS differentiates between food eaten within and outside the household, allowing some implications to be made. Nonetheless, the data are not robust enough to allow comments upon (and therefore comparisons with this study) with regard to eating of processed foods. If this high percentage of energy from processed food is typical of a GFD, appropriate and detailed labelling of processed food is vital so as those on a GFD can make fully informed decisions.
In summary, this study of nutrient intakes in CD reveals adequate macronutrient intakes, but higher proportions of carbohydrate from NMES. Intakes of magnesium, iron, zinc, manganese, selenium and folate in women are lower than a comparative population taking a gluten containing diet. In men, the percentage of patients with CD achieving the RNI for magnesium and selenium is particularly low. While mean intakes of calcium are higher than in gluten containing diets, many patients, particularly those over 55 years of age, fail to achieve BSG-recommended intakes for calcium.
Current dietary advice and recommendations for supplementation need review to take account of these deficiencies. Where possible, recommendations should be made to address the low intakes of fibre, calcium and iron. Maintaining available carbohydrate intakes without resorting to sugary foods should be done. Patients should be encouraged that this could occur without the need to wholly rely on special GF products, but rather by increasing intake of food products such as beans and pulses, starchy vegetables such as sweet potato and potato or alternative grains as recently suggested by Lee et al.26
This work highlights the importance of all patients with coeliac disease having the opportunity for dietary review with a suitably experienced dietician on a regular basis for advice on foods to achieve a balanced diet or appropriate dietary supplementation.
Declaration of personal and funding interests: None.