Systematic review: adherence to a gluten-free diet in adult patients with coeliac disease

Authors


Prof. G. Rubin, School of Medicine and Health, Durham University, Wolfson Research Institute, Queen’s Campus, University Boulevard, Stockton on Tees, TS17 6BH, UK.
E-mail: g.p.rubin@durham.ac.uk

Summary

Background  Coeliac disease is increasingly diagnosed in adult patients who present with atypical symptoms or who are asymptomatic and detected by case screening. Its treatment, a gluten-free diet, can have a considerable impact on daily living. Understanding the factors associated with non-adherence is important in terms of supporting patients with their condition.

Aim  To investigate factors associated with adherence to a gluten-free diet in adults with coeliac disease.

Methods  A literature search of multiple electronic databases using a pre-determined search string for literature between 1980 and November 2007 identified a possible 611 hits. After checking for relevance, 38 studies were included in this review.

Results  Rates for strict adherence range from 42% to 91% depending on definition and method of assessment and are the lowest among ethnic minorities and those diagnosed in childhood. Adherence is most strongly associated with cognitive, emotional and socio-cultural influences, membership of an advocacy group and regular dietetic follow-up. Screen and symptom-detected coeliac patients do not differ in their adherence to a gluten-free diet.

Conclusions  The existing evidence for factors associated with non-adherence to a gluten-free diet is of variable quality. Further and more rigorous research is needed to characterize those individuals most likely to be non-adherent to assist them better with their treatment.

Background

Coeliac disease (CD) is chronic inflammatory intestinal disorder induced by gluten ingestion in genetically susceptible individuals. It affects 0.5–1% of the population, two thirds of whom remain undetected1–5 and is increasingly first diagnosed in adulthood.4 The availability of serological testing has seen a rise in individuals being diagnosed with atypical, minimal or no symptoms.6 These tests are being used for active case-finding7, 8 and for targeted screening in high risk groups.9–11

The mainstay of treatment for CD is life-long adherence to a gluten-free diet (GFD). For the majority of patients, the introduction of a GFD results in full clinical and histological remission and has been associated with improvements in quality of life12 and a decrease in the long-term health risks of osteoporosis,13 gastrointestinal malignancies,14–16 nutritional deficiency and the development of autoimmune disorders such as diabetes mellitus. Adherence to a GFD can also increase the likelihood of health gains in problems associated with CD such as infertility,17 fatigue18 and depression.19–21

Reported levels of strict adherence to the GFD in adults with coeliac disease vary between 36% and 96%.13, 22 Compliance to treatment in chronic gastrointestinal disease is generally high (mean 80.4%) compared to other chronic illnesses including cardiovascular diseases (mean 76.6%) and diabetes (mean 67.5%).23 Adherence to dietary advice, however, is reported to be the lowest of all treatment types.23 For patients with CD, adherence to a GFD can be restrictive, difficult to follow and impinge upon their quality of life.24 Despite gluten-free products being available on prescription in the UK and other European countries, an improvement in their general availability over the last few years and recent changes in food allergen labelling regulations both in Europe24 and the US, exclusion of gluten-containing foods is made difficult by confusing food labelling, limited availability and the high cost of gluten-free alternatives.25

Treatment complexity, uncertainty of treatment options and chronic illness have all been associated with lower adherence to treatment advice,23, 26 along with psychosocial factors including illness and treatment beliefs27, 28 and depression.29 Current evidence relating these and other factors to adherence in CD22, 30–36 is often contradictory and no systematic review has been carried out to date. Furthermore, there is little information on the use of interventions in improving adherence in CD.37 The aim of this review was, therefore, to identify the factors associated with adherence to a GFD in adults with CD and to determine the effectiveness of any interventions.

Methods

We searched Embase, Medline, Pub Med, Psychinfo, Cinahl and the Cochrane library for literature from 1980 to November 2007 using the following search terms: (coeliac or celiac or ‘gluten sensitive enteropathy’) and [(diet$ or nutrition$ or GFD or ‘gluten-free’ or ‘gluten free’) with (advice or adherence or compliance or concordance or prescription or intervention or management)]. Papers were excluded if they combined data on children and adults or if coeliac disease was not the primary illness under study.

Papers eligible for inclusion were assessed to ensure that they reached a minimum quality standard, based on previously defined criteria.38, 39 Data were extracted using a standardized form. Reference lists of retrieved papers were scrutinized for additional papers.

The search criteria identified 611 potential papers which were reduced to 401 after deleting papers relating to children only, those published in a language other than English and those not primary studies. Abstracts were reviewed by NH and GR to identify those relevant to the aims of the review, for which the full papers were then obtained. After additional hand searches of identified papers, 38 studies fulfilled the inclusion criteria for this review. For reasons of the range of largely descriptive methodologies employed, it was not possible to apply a single quality assessment method to studies.

Results

Table 1 summarizes 10 studies22, 30, 40–47 which have as their main aim examining the factors associated with adherence in CD directly or in association with quality of life. Fourteen additional studies21,33, 48–59 (Table 2) report on factors associated with adherence as secondary to their main aims and a further 146, 10, 20, 32, 60–70 provide descriptive accounts of the issues surrounding adherence to the GFD (Supplementary material available on-line).

Table 1.   Summary of studies examining factors associated with adherence in CD
AuthorsDateCountrySample and study designAge range or mean (years)Primary aim of researchAdherence measured byFactors not associated with adherenceFactors associated with adherenceComments/Results
Addoloratto et al.402004ItalyClinical sample (58.9% anxiety or depression) Randomized controlled trial
= 33 (treatment group)
= 33
 (control group)
Treatment group: 31.6 (s.d. = 5.8)
Control group:
29.8 (s.d. = 7.2)
To assess impact of psychological support counselling on adherence and affective disorders in CD with anxiety or depressionSelf report, family member interview, serology, histology and clinical assessment×DepressionNo differences found in state anxiety between groups at 6 month follow-up. Lower percentage of depression in treatment group (= 0.001). Lower compliance to GFD was found in control group (= 0.02).
Butterworth et al.292004UK Clinical sample Cross-sectional survey with case note review = 66 (White Caucasian) = 21 (South Asian)>16To investigate factors relating to compliance in White Caucasians & South AsiansQuestionnaireNone of the factors associated with adherence in South AsiansCoeliac society membership, understanding food labelling, obtaining sufficient gluten-free products, explanation by physician, regular dietetic follow-upFactors associated with adherence were only significant in White Caucasian group. Other differences between the groups found with regard to follow-up, membership of coeliac societies and satisfaction with care. Pre-defined list of factors used.
Casellas et al.412006SpainClinical sample Cross- sectional survey
= 62 established CD
= 11 newly diagnosed
>20 years
Median 38 [30–55]
To provide info on heath care seeking behaviour, info sources & treatment issuesAdapted MMASAge, education level, duration of GFD difficulty of diet, symptoms after gluten×Difficulty with the diet was lower and number of patients who experienced symptoms after gluten was higher in compliant group, although these differences did not reach significance. Unintentional lapses were reported by 15.5% of patients and intentional lapses by 1.8%.
Goggins et al.42
(conference abstract only)
1995UKClinical cohort
Cross-sectional study
(85% response rate)
= 91
20–80Audit of complianceQuestionnaireTime since diagnosis or SES×42% reported not adhering strictly to GFD (10% often included gluten) Cost, problems with social occasions, difficulty finding GF products reported to contribute to difficulties in adhering.
Gremigni et al.43
(conference abstract only)
2006ItalyClinical cohort
 Case–control study
= 67 CD
= 61 matched controls
18–62To examine relationship between illness representations, well-being and adherence to the diet in adult coeliac patientsQuestionnaire (4 items)×Gender
Illness representations
Men less compliant than women. Specific dimensions of illness representations and burden of illness accounted for 15% of variance in adherence, 9–30% well-being and 6–10% QOL.
Hogberg et al.222003 SwedenConsecutive clinical sample
Retrospective cohort study All diagnosed under age of 18 years. = 15 < 4 years diagnosis
= 14 > 4 years diagnosis
19–26Investigate if diagnosis at <4 years = better compliance at follow-up (17–24 years)Interview/questionnaire/serology/histology ×Age at diagnosis80% of CD patients diagnosed <4 years were compliant, 36% of CD patients diagnosed >4 years were compliant (< 0.05)
Kokkonen et al.441998FinlandClinical cohort (93% of former patients- all childhood diagnosis) Cross-sectional survey = 4217–26To assess degree of compliance in young adultsQuestionnaireAge at diagnosis, social class, education or SES×64% continue strict diet. Those adhering had good knowledge of CD. Only 3 patients had low knowledge about the diet. These were all noncompliers, although no significant differences in knowledge was found between the two groups. Quality of paper relatively poor.
Lamontagne et al.452001QuebecQuebec Celiac Foundation Cross-sectional survey (ran dom sample, 48% response  rate) = 234>18To analyse dietary habits and related problemsQuestionnaireGender, education, health problems, satisfaction with price or labelling of GF products, Frequency of reading food labels or eating out, social worriesAge, satisfaction with taste/texture of GF products, hi confidence in treatment information, size of region of residence, worry over food preparationRange of factors described associated with difficulty in complying with the GFD rather than compliance per se. Lack of confidence in treatment information was most strongly related to difficulty in complying with GFD (< 0.005). 90% avoided gluten as much as possible. 20 questionnaires disregarded as not on GFD. Younger respondents and those living in cities had the greatest difficulty complying.
Leffler et al.462007USA28% clinical sample
72% other
Cross sectional study
= 154
22–91To identify factors independently correlated with GFD adherenceQuestionnaire,
assessment by expert nutritionist and serological testing
Gender, age, age at diagnosis, time on GFD, educational achievement, employment status, presence of symptoms at diagnosis, presence of psychological disturbance, sample type, perception of availability and quality of GF foods, satisfaction and information from health care providerMarital status,
Presence of additional food intolerances, cost of GF food, ability to avoid foods away from home, ability to avoid foods irrespective or mood or stress, self reported and actual understanding of GFD, beliefs about harm from exposure to gluten, membership of CD advocacy group, belief that avoiding gluten is important for health
GCAS formulated by range of experts and patient focus group. A range of psychosocial factors were found to be associated with adherence. Age, age at diagnosis, time on GFD, gender, education and marital status were controlled for.
Viljamaa et al.472005 FinlandSecondary care sample (consecutive screen detected and random symptom detected)
Case–control study
= 53 screen detected
= 44 symptom detected
= 110 non-CD
20–75To investigate compliance, QOL and bone mineral density in screen detected CD. 14 year follow-upDietetic interview,
 4 day food record, serological tests
Age at diagnosis, duration of follow-up, age, family history, GI symptoms, QOL×Long-term diet compliance is good in both screen (96% strict or fairly strict) and symptom detected CD (93% strict or fairly strict). (P = NS) QOL and GI symptoms in screen detected CD similar to symptom detected CD and non CD controls.
Table 2.   Studies which have examined factors associated with adherence as secondary to main aims of research
AuthorsDateCountrySample and research designAge range or mean (years)Primary aim of researchAdherence measured byFactors NOT associated with adherenceFactors associated with adherenceComments/Results
Casellas et al.482005SpainConsecutive clinical sample – cross- sectional survey
= 54 on GFD
= 9 newly diagnosed
GFD group: Median 35 [30–54]
 Newly diagnosed group 41 [26–54]
To evaluate perceived health status in CDAdaptation of MMASAge, treatment duration, education, QOL×Perceived easiness of GFD was consistent with compliance, although no significance levels are reported. QOL was found to be higher in treated CD
Ciacci et al.491998ItalySecondary care consecutive sample – case–control study
 n = 92 CD
 n = 100 health controls
= 48 hepatitis
20.56 (s.d. = 15.35)To explore relevance of depressive symptoms in adult CDSelf-report (Visual analogue scale)Age, gender, socioeconomic status depression×Depressive symptoms feature of disease and not related to adherence to GFD. Knowledge was significantly associated with higher socioeconomic status (< 0.05)
Ciacci et al.502002aItalySecondary care consecutive sample – cross sectional survey
= 114 CD on GFD
= 25 CD not yet on GFD
29.62 mean (s.d. 11.18)To evaluate emotional impact of CD diagnosis in adulthood, relationship between patients and doctors and how patients cope emotionally with disease and dietSelf-report questionnaire (Visual analogue scale)Past adherenceAnger
 duration of illness
Anger was inversely correlated to compliance with the GFD (= 0.0005) and the predominant emotion that induced patients to transgress. A significant difference was found between past and present adherence (= 0.0025). Patients of higher socioeconomic status showed better knowledge of disease (= 0.001)
Ciacci et al.512002bItalySecondary care prospective cohort study= 39027.9 mean (s.d. 10.9)Long term follow-up of adult CD: prevalence and correlates of intestinal damageInterview by trained physicianGender, occupationIntestinal damage; duration of follow-up, baseline blood Hb; baseline age; educationLack of adherence to strict gluten-free diet is main reason for poorly controlled disease in adults. Negative correlation with baseline blood hb (= 0.021) and duration of follow-up (= 0.013) suggests that those with more severe symptoms of malabsorption at diagnosis have better compliance. Baseline age, which was taken at diagnosis (< 0.05) and education (< 0.05) were positively correlated to adherence.
Ciacci  et al.522003ItalyCoeliac society members
Cross-sectional survey
(78% response rate)
= 581
31.5 mean (s.d. 11.3)To evaluate quality of life in adult CDSelf-report (visual analogue scale)Gender,
anxiety
Age at diagnosis,
Educational level,
Uneasiness and embarrassment at sharing table
Compliance was correlated to age at diagnosis. Those diagnosed after the age of 20 years and with higher education levels had better compliance. (= 0.0001). Amount of discomfort and embarrassment was negatively related to adherence (= 0.0001) Gender differences in compliance (= 0.0025) became insignificant once adjusted for age at diagnosis. Anxiety scores were not correlated to age at diagnosis or gender. Reasons for noncompliance included problems in ordering in restaurants, a feeling of anger towards CD, not to be different from others, hope that occasional gluten is not harmful.
DeRosa  et al.532004ItalyConsecutive clinical sample – case–control study = 29 CD = 47 (healthy controls)Mean 26.72 (s.d. 9.43)To evaluate illness behaviour and examine influence of CD and GFD on personality and adherenceSelf-report (Visual analogue scale)×Total duration of disease
age of patient
Findings suggest that CD may be associated with changes in personality that may interfere with the patients’ adaptation to living with a chronic disease. Adherence was positively related to total duration of disease (= 0.02) and to current age (= 0.009)
Fera et al.542003ItalyConsecutive clinical sample – case–control study
= 100 CD
 (GFD>1 year)
= 100 healthy controls
= 100 diabetes
20–70To estimate incidence of affective disorders in CDdietetic assessment & self reportAge, age at diagnosis, diagnostic delay, clinical presentation, duration of GFD, depression or anxiety, QOLHistological outcome and EMAAnxiety and depression higher in CD than diabetic controls. QOL poorer in CD and diabetics than in controls. Self-reports for adherence were higher than expert assessment
Hallert et al.551998SwedenClinical cohort (GFD >10 year) (inclusion rate 87%) cross-sectional survey = 8935–74To measure perceived health status, correlate to histologySelf-reportQOL  Women with CD had lower QOL scores than general population controls. Separate male and female controls were used. Patients reporting less compliance with GFD did not have lower QOL scores
Hauser et al.562007GermanyGerman Coeliac Society Cross-sectional survey (52.2% response rate) = 446 18–88Predictors of reduced HRQOL in CDSelf-report×QOL (CDQ only)Non-adherence predicted reduced QOL in the CDQ only and not the SF-36. Physical and mental co-morbidities and younger age also associated with reduced QOL
Usai et al.572002ItalyConsecutive clinical sample (GFD>2 years) Case–control study = 68 CD = 136 healthy controls18–74To evaluate QOL in relation to severity of illness, compliance and associated diseasesSerology and self- report×QOLQOL in CD was lower than in healthy controls. Compliers had better scores than noncompliers on the mental health and social functioning domains (> 0.05). Severity of illness and associated disease also associated with QOL
Usai et al.582007ItalyConsecutive clinical sample (GFD>2 years) Case–control study = 129 CD = 526 matched non-IBS controls = 102 IBS controls18–74To evaluate effect of GFD adherence and IBS symptoms on QOL in adultSerology and self- report×QOLIBS and CD had lower QOL scores than healthy controls (> 0.05). CD strict compliers showed better scores than partial compliers in all domains except physical functioning, physical-role and bodily pain (< 0.05). Lowest scores were found in partial diet CD with IBS symptoms
Vahedi et al.332003France/ BelgiumConsecutive secondary care sample Cross-sectional study = 9517–74To assess the reliability of anti-transglutaminase antibodies as predictors of gluten-free diet compliance in adult CDSerology; dietician interview; biopsyGFD duration; age×Transgressions were reported to bevoluntary in 69% of noncompliant patients. Reasons for voluntary noncompliance provided include poor palatability, symptoms after transgressions, constraints greater than benefits, high cost and cultural reasons. Involuntary transgressions because of nonspecified presence, erroneous affirmation of gluten absence, gluten containing drugs. No differences found in age or GFD duration between the compliant and noncompliant groups
Whitaker et al.
 (conference abstract) 34
2004UKClinical sample Cross-sectional survey (83% response rate) = 108 symptomatic; = 62 asymtpomaticNRTo assess burden of illness and GFD and compare between symptomatic and asymptomatic CDquestionnaire××Describes some of the difficulties experienced by CD patients. More asymptomatic CD patients (27%) expressed regret at diagnosis than symptomatic CD (10%) (< 0.01)
Wylie et al.592005UKSecondary care patients attending coeliac dietitian led clinic Prospective cohort design = 9923–86To follow-up clinical audit of patients attending the dietitian-led coeliac clinicSelf-report and assessment by dietitian using food diary and interview×Attendance at coeliac clinic100% satisfaction with appointment and services offered. Significant increase in actual dietary adherence (< 0.05) as well as other dietary markers at year 1 compared to baseline

Definitions of adherence

The studies identified for this review used a variety of definitions of adherence intrinsically tied to the way in which adherence was assessed and measured. Where self-report or dietary interview were used, for example, adherence was often defined as a discrete variable i.e. strict, partially or fairly strict, and non-adherent. Some studies considered strict or fairly strict adherers to be ‘adherent’, whereas others excluded those who were ‘fairly strict’. These categories were either defined prescriptively, e.g. as ‘less than one serving of gluten per week’ or an estimate of milligrams of gluten ingested, or else by vague statements such as ‘adheres to the diet most of the time’. Visual analogue or Likert scales were used as continuous variables. In some instances, adherence was defined in relation to histopathology or serological tests. For reasons of mainly observational designs of the studies and varied definitions and measurement of adherence, it was not appropriate to carry out a meta-analysis.

The studies included in this review report strict adherence rates as measured by expert assessment ranging from 44% to 90% and by self-report from 42% to 91%. Comparing adherence levels across studies is problematic, however, because of the large variation in measurement and definitions of adherence. Complete non-adherence rates vary from 0% to 32% although most are below 5%.23, 26 The highest non-adherence rates were found in studies based on populations diagnosed in childhood44, 66 and ethnic minority groups.30

Sociodemographic factors

The weight of the available evidence is against a correlation between adherence and either education or socioeconomic status41, 42, 44, 46, 48, 49 with the notable exceptions of two large studies, both from the same Italian group, which found education to be significantly correlated to adherence.51, 52 The balance of evidence is also against an association between current age and adherence in adults with CD,33, 41, 46, 47, 49, 54 with only two studies reporting a positive association.45, 53 The effect of age at diagnosis is less clear. Some studies report no association,46, 47, 54 whereas others have shown that younger age at diagnosis is related to lower adherence51, 52 and a UK study on patients diagnosed in their 7th decade shows that adherence is relatively high in patients diagnosed later in life.63 For adults diagnosed in childhood, diagnosis under the age of 4 years has been related to better adherence at follow-up.22 Across all the studies identified, there is a trend for non-adherence rates to be higher among adults diagnosed in childhood than in those diagnosed in adulthood. Some limited explanation may be provided by a qualitative study which describes differences in the worries and dilemmas about living with CD faced by the informants diagnosed as children compared to those diagnosed as adults.67

Although gender has been shown to influence quality of life, burden of illness and the presence of ongoing symptoms in CD,20, 55, 61, 68, 70 there is no consistent association between gender and adherence.43, 46, 49, 51, 52

Knowledge, attitudes and beliefs

Several studies offer useful descriptive accounts and explanations of the illness and treatment-related attitudes and beliefs associated with having to follow a GFD,10, 20, 32, 42, 48, 50, 61, 62, 66–70 although only a few have attempted to measure the significance of the association between these variables and adherence. Adherence has been found to be significantly associated with illness representations as measured by the IPQ-R,43 embarrassment or uneasiness at table sharing and feelings of anger towards the illness52 and beliefs about the harmful effects of gluten or concerns about exposure.46, 52

Although an association between poor reported and actual knowledge about the GFD and noncompliance has been identified,46 noncompliance does not necessarily imply poor knowledge. One study, for example, found that despite all the patients with poor knowledge being noncompliant, not all of the noncompliers had poor knowledge and no significant difference in knowledge was found between compliers and noncompliers. The number of patients with poor knowledge of the GFD, however, was very low.44

Illness and symptom factors

There are no differences in levels of adherence to a GFD between screen detected and symptom detected patients with CD,6, 47, 64 and no association has been found with pattern of clinical presentation54 or symptoms prior to diagnosis.46 One prospective cohort study looking at symptoms at diagnosis rather than clinical presentation, however, identified a significant negative correlation between compliance and blood haemoglobin measured at baseline, suggesting that participants with more severe symptoms and signs of malabsorption before treatment were more likely to have better dietary compliance upon follow-up.51

Despite incidental evidence identifying the absence of symptoms as one of the main reasons for voluntary transgressions,33, 66 no study has confirmed a statistically significant association between the presence of symptoms with gluten ingestion after diagnosis and adherence. This association could, however, be bi-directional, as a lack of symptoms may be associated with high adherence in those who have had symptoms in the past and are motivated to comply strictly, yet may also be associated with low adherence in those who are not motivated and do not experience symptoms after gluten ingestion. Furthermore, the association with adherence may be complicated by the wide range and severity of symptoms experienced by individuals.

Although cross-sectional studies have found that adherence improves with illness duration,50, 53 one prospective cohort study found that it declined,51 and other studies have found no differences in GFD duration between compliant and noncompliant groups.33, 41, 46–48, 54

Treatment factors

The GFD is widely reported to be restrictive, complex, costly and difficult to follow.10, 21, 33, 42, 44, 45, 66, 69, 70 Although this is generally acknowledged as an important issue, there is little evidence that these factors are consistently significantly associated with adherence,41, 45, 46 One UK based study, nevertheless, found that understanding food labelling, affordability, obtaining GF foods and obtaining enough GF foods on prescription were all significantly associated with adherence in White Caucasians, although these associations were not significant in a South Asian sub-set.30

Patients with CD have variable confidence in the advice given by health professionals and place greatest store in the information provided by dieticians.42, 44–46, 48, 69 Lack of confidence in treatment information from the gastroenterologist and dietician was associated with difficulty in adhering to the GFD in one study,45 although others have found that satisfaction with information and support provided by healthcare providers was not related to adherence.46, 50

One published clinical audit concluded that annual review within the context of a dietician–led coeliac clinic can significantly improve adherence as well as other nutritional markers.59 Regular dietetic follow-up and detailed discussion of the disease by the physician have also been found to be associated with adherence.30

Membership of a patient support group appears to be associated with adherence.30, 46 In general, studies that have recruited from patient support groups have shown a trend towards higher adherence rates (66–90% strict adherence) than clinical samples (42–91% strict adherence). The percentage of patients who are members of coeliac societies in studies that sample clinical populations is not always known and without this information, it is difficult to draw any firm conclusions about the impact of membership.

Socio-cultural/Environmental factors

Although no relationship has been identified between adherence and socioeconomic status,41, 42, 44, 46, 48 there is some limited evidence of an effect for other sociocultural factors. Reported ability to follow a GFD when travelling, dining out, at work and during social events has been found to be significantly related to adherence.46 In addition, differences in terms of GFD adherence have been described between white Caucasian and South Asian groups suggesting cultural differences in knowledge and beliefs about the diet.30

Country or region of residence, in particular in terms of public awareness, the availability of gluten-free foods and compatibility with usual lifestyles, has been posited as an important variable in explaining study findings,45, 47, 51 although this has not been specifically examined.

Quality of life and psychological well-being

Levels of anxiety and depression are increased in coeliac disease,40, 49, 50, 54, 71, 72 although there is debate over whether they are part of the ‘coeliac pattern’ or whether they occur as a result of coping with a chronic illness.19, 54, 71, 72 No correlation between self-reported compliance and depressive symptoms was found in a questionnaire survey49 and in a large case-controlled cohort study.54 One randomized control trial has shown that if depressive symptoms improve, so does adherence.40 Anger and reported ability to follow the diet despite changes in mood or stress have also been found to be correlated with adherence.46, 50

Some studies have found that HRQOL, although improved after starting a gluten-free diet, may still be reduced in comparison with that in general population.18, 54, 56, 58, 62 Nevertheless, others have found no significant differences in QOL or psychological well-being between CD and population controls at short-term and long-term10, 47, 48, 55, 69, 70, 73 and for certain cohorts of patients, quality of life is reported to be higher than in the general population.6, 66 Adherence to the diet is not always accounted for in these studies, while different measures and definitions of quality of life as well as adherence limit their comparability. Overall, the evidence suggests that compliance is not related to quality of life,47, 48, 54, 55 although three studies have reported associations within certain quality of life domains.56–58

Interventions

Only two studies report on interventions to improve adherence to a GFD in adults. In the first, sixty-six newly diagnosed patients with either anxiety or depression were randomized to a psychological support treatment group or a control group and followed up every 2 weeks for 6 months. Although no differences were found in levels of state anxiety between the groups, the intervention group had a lower percentage of depressed subjects at follow-up. Adherence to GFD was assessed by the participants’ self-report, a family member interview, clinical symptoms and histological recovery and by disappearance of AGA and EMA. Three (9.1%) patients in the intervention group showed minor lapses with their GFD, whereas 13 (39.4%) of patients in the control group were non-adherent, 5 (12.1%) of whom had minor lapses.40

A second randomized controlled trial compared existing training offered by the German coeliac society with a computer-based interactive training programme. The intervention group had significantly higher recall of acquired knowledge and improved ability to transfer knowledge about the GFD into other situations at post-treatment and at follow-up. Actual adherence and symptom or histological improvements, however, were not assessed.65

Discussion

Although complete non-adherence to GFD is uncommon among patients with CD, up to 60% are partially non-adherent. Studies have examined a range of demographic, illness, treatment, psychosocial and cultural factors related to adherence to the GFD. The factors most strongly associated with adherence are cognitive, emotional and sociocultural influences, membership of an advocacy group and regular dietetic follow-up. The evidence to support these conclusions is, however, limited. Although there is no strong evidence to support a difference in terms of adherence between screen- and symptom-detected patients, severity of malabsorption before treatment may be related to better adherence. There is weak evidence for a relationship between adherence and the practical difficulties associated with the complexity of the GFD, accessibility of gluten-free alternatives and education level.

The main limitation of this review is the lack of comparability between studies, in terms of design, methods, definitions and measurement of adherence. There is no agreed gold standard for measuring or monitoring adherence. Histology takes up to 12 months to return to normal after commencing a gluten-free diet74 and repeat biopsies are not always an acceptable monitoring option. Serological tests are not sufficiently sensitive to detect minor transgressions to the diet and their association with histological results has been challenged.33, 75–79 Dietetic assessment by an expert, usually based on an interview or food diary is considered the most objective non-invasive method of measuring adherence.46, 51 Moreover, the lack of agreement over what constitutes a ‘strict’ GFD corresponds to an important clinical issue, as the concept of a safe gluten threshold80 and variation in individual tolerance to trace levels of gluten is the subject of debate.74, 81–86 This is reflected in the development of a new dual codex standard for gluten-free, adopted by the EU in January 2009, which reduces the amount of gluten allowed in foods labelled as ‘gluten-free’ from 200 to 20 mg parts per million24 and in the recent definition of ‘gluten-free’ by the US Food and Drug Administration.87

Most studies have recruited participants from secondary care clinics or members of coeliac associations, which can both be important sources of support and information for those who have access, or choose to access them.31 The quality of this support and its impact on adherence may vary, however, and is difficult to account for. In some countries, moreover, members of these associations are not required to have a biopsy-confirmed diagnosis. At the same time, those studies that recruited through coeliac associations tended to have larger sample sizes and would be less likely to be subject to context-specific influences, such as geographical area or hospital clinic.

People who do not adhere may be less likely to return questionnaires or agree to participate in the research, resulting in a general over-estimation of adherence rates. The majority of published studies are observational and, even when correlations are significant, it is not possible to make any conclusions about causality or the potential for improvement. Lastly, intentional non-adherence and non-intentional non-adherence are rarely distinguished, and may be moderated in different ways by different factors. Good knowledge of the GFD, for example, may be related to higher levels of self-reported intentional non-adherence, whereas those with poor knowledge may not be aware of their gluten ingestion and are likely to report lapses as unintentional, if noticed at all. Defining and measuring both types of adherence would be necessary to examine any such effects.

This systematic review sought to identify factors influencing adherence to GFD in adults with CD. As such, it did not examine the effect of differing levels of adherence on long-term clinical outcomes for these patients.

Future studies of adherence in coeliac disease should be methodologically robust, paying particular attention to sampling, the terminology and measurement of adherence and the use of validated survey instruments. Further research is needed to help understand the circumstances in which patients deviate from the GFD and to develop and test appropriate and effective strategies to assist those who need support with their treatment.

Acknowledgements

Declaration of personal interests: None. Guarantor of the article: NH and GR accept full responsibility for the conduct of the study and the decision to publish. NH was responsible for the study conception and design, acquisition of data, reviewing abstracts, interpreting findings and the drafting of manuscript. GR assisted with the study conception and design, reviewed abstracts and contributed to the drafting of the manuscript and interpretation of findings. AC was involved in the study conception and design and reviewed the manuscript. All authors approved the final draft submitted. Declaration of funding interests: This paper was completed as part of a PhD programme of study funded by the Northern Primary Care Research Network and the University of Sunderland.

Ancillary