Evidence for the recognition of coeliac disease

4Evidence for the recognition of coeliac disease

4.1. Signs & symptoms

4.1.1. Review question

What are the clinical signs and symptoms that raise suspicion of coeliac disease?

4.1.2. Methods

The aim of this review question was to establish which presenting clinical features may raise suspicion about the presence of coeliac disease and the need for serological testing. This is an update of the chapter on ‘clinical signs and symptoms of coeliac disease’ in the 2009 guideline for coeliac disease (CG86). This updated review incorporates studies that were included in the previous guideline together with newly-published evidence.

Studies were considered if they met the following inclusion criteria: the population examined was children, young people, or adults with undiagnosed coeliac disease who presented with clinical signs and symptoms that may raise suspicion for the disease; coeliac disease diagnosis was confirmed by intestinal biopsy. Studies were excluded from analyses if people were receiving treatment for coeliac disease at the time of testing; if the diagnosis of coeliac disease was not confirmed by intestinal biopsy; if the population of interest had non-coeliac gluten sensitivity or wheat allergy.

An exhaustive list of clinical signs and symptoms was suggested by the GDG within the review protocol (see Appendix C) prior to conducting the literature searches for this review question. The comparator was a biopsy-confirmed diagnosis of coeliac disease. The outcome of interest was the diagnostic utility of the odds ratio of having coeliac disease, given the presence of a particular sign or symptom, compared to the odds of not having the disease in the absence of that sign or symptom.

Case-control studies were considered the most appropriate study design to derive signs and symptoms odds ratios for this question and were therefore considered the highest quality within a modified GRADE framework. All other study designs were downgraded from this review, including cohort studies, case-reports, case series, and qualitative studies. Studies could be downgraded due to reasons such as imprecision of odds ratio metrics, the presence of study bias, or inconsistency of effect estimates between studies.

Included studies

A single systematic search was conducted for sections 4.1, 4.2, and 4.3 (see Appendix C) which identified 7230 references. This search was restricted to studies published from 2008 onwards to avoid duplicates of studies considered in the previous coeliac disease guideline (CG86). The references were screened on their titles and abstracts and full papers of 134 references were obtained and reviewed against the inclusion and exclusion criteria in the review protocol (see Appendix C).

Overall, 128 studies were excluded as they did not meet the eligibility criteria such as inappropriate study design (case-series), not a primary study (descriptive narrative, opinion, etc.), examined the prevalence of coeliac disease in certain populations, studies in which the study population was not suspected of coeliac disease (but may have had an increased risk for developing coeliac disease, such as a commonly comorbid condition, or a family history of coeliac disease), and studies which did not use Marsh grade 3 for the histological diagnosis of coeliac disease. A detailed list of excluded studies and reasons for their exclusion is provided in Appendix F.

The 6 remaining published papers did meet the eligibility criteria and were included. Data was extracted into detailed evidence tables (see Appendix D) and are summarised below. Two published papers (Olen et al., 2008; Mollazedegen et al., 2009) utilised the same database of Swedish men who underwent medical testing prior to and post conscription. Three published papers (Olen et al., 2008; Mollazedegen et al., 2009; Ludvigsson et al., 2013) also utilised the same control database of Swedish national register of patient information.

The 11 included studies in the previous coeliac disease guideline (CG86) were reviewed against the current protocol. Of these, all were excluded as they did not meet eligibility criteria. Primary reasons for exclusion included inappropriate study design, such as prevalence studies; populations that were already being treated for coeliac disease at the time of inclusion, or who were examined for a sign or symptom after the diagnosis of coeliac disease, and studies in which coeliac disease diagnosis was not biopsy-confirmed.

The overall quality of the evidence from these published papers ranged from very low to moderate, with the majority of evidence to be of low quality.

4.1.3. Evidence review

4.1.3.1. Intussusception in adults

One large nation-wide population study (Ludvigsson et al., 2013) examined the relationship between intussusception and coeliac disease in a group of newly-diagnosed coeliac disease patients (n=29096; mean age 30 years) and healthy control participants (n=144522; age-matched). The prevalence of intussusception was very low in both the coeliac (0.12%) and control (0.10%) groups.

4.1.3.2. Low BMI in adults

One large study (Olen et al., 2009) identified adult males with or without coeliac disease through a national inpatients register of men who were admitted to hospital immediately prior to or post conscription. This study was broken down into two parts; a cohort study and case-control study. Only case-control study data examining the association between low BMI and coeliac disease was used in this analysis. In order to be eligible for the study, data on weight and height had to have been available at the time of study. Only data for those diagnosed with coeliac disease after review of weight and BMI were included in the analyses (n=70; age range 18–50 years). Controls were identified via a government total population register and were matched to patients for age, sex, and country of residence (n=6887; age range 18–50 years).

4.1.3.3. Impaired visual acuity in adults

One large study (Mollazadegan et al., 2009) identified adult males with or without coeliac disease through a national inpatients register of men who were admitted to hospital immediately prior to or post conscription. This study was broken down into two parts; a cohort study and case-control study. Only case-control study data examining the association between impaired visual acuity and coeliac disease was used in this analysis. In order to be eligible for the study, data on visual acuity had to have been available at the time of study. Only data for those diagnosed with coeliac disease after review of visual acuity were included in the analyses (n=69; mean age 18.9 years). Controls were identified via a government total population register and were matched to patients for age, sex, and country of residence (n=6850; mean age 18.7 years).

4.1.3.4. Migraine in children and young adults

Two studies were identified (Alehan et al., 2008; Inaloo et al., 2011) that examined the association between migraine and coeliac disease. Alehan and colleagues (2008) examined the prevalence of coeliac disease in a population of paediatric patients with migraine (n=73; mean age 12.01 years) and healthy control participants (n=147; mean age 11.85 years). One case (0.7%) was identified in the control group, and 4 cases (5.55%) with coeliac disease were identified in the migraine group. Inaloo and colleagues (2011) similarly examined the prevalence of coeliac disease in a population of paediatric migraine patients (n=100; mean age 9.5 years) and healthy control participants (n=1500; mean age 10.6 years). Equal prevalence estimates (2%) of coeliac disease were reported for both the patient and control groups.

4.1.3.5. Dental enamel defects in children and young adults

One study (El-Hodhod et al., 2012) examined the frequency of coeliac disease in paediatric patients with dental enamel defects (n=140; mean age 8.33 years) and healthy controls (n=720; age range 4–12 years). The control children were recruited as part of their routine annual health check in the local children’s hospital as part of the well child clinic. All participants underwent an oral hygiene and dental examination and IgA and IgG tTG serological testing for coeliac disease. Positive serological results were followed-up with an endoscopic intestinal biopsy to confirm or exclude the presence of coeliac disease.

4.1.4. Health economic evidence

An economic literature search was not conducted for this question as an economic evaluation would not be the correct framework in which to generate useful evidence on the signs and symptoms of coeliac disease.

4.1.5. Evidence statements

4.1.5.1. Evidence for the relationship between intussusception and coeliac disease in adults

One very low quality published paper (Ludvigsson et al., 2013) of 173618 adults reported no association between coeliac disease and intussusception (OR 1.18, 95% CI: 0.81 to 1.71).

4.1.5.2. Evidence for the relationship between low BMI and coeliac disease in adults

One very low quality study (Olen et al., 2009) of 6957 adult males reported a trend toward an association between low BMI (<18.5) and coeliac disease compared to a healthy age and gender matched population (OR 2.2, 95% CI: 1.0 to 4.8); however as the lower confidence interval lies on 1.0, this was not statistically significant. .

4.1.5.3. Evidence for the relationship between visual acuity and coeliac disease in adults

One very low quality published paper (Mollazedagen et al., 2009) of 6957 adults males reported no associated between impaired visual acuity and coeliac disease compared to a healthy age and gender matched population (OR 1.04, 95% CI: 0.9 to 1.19).

4.1.5.4. Evidence for the relationship between migraine and coeliac disease in children and young adults

One very low quality study (Alehan et al., 2008) of 220 children and adolescents reported an association between migraine and coeliac disease compared to a healthy age and gender matched population (OR 8.46, 95% CI: 0.92 to 77.15); however this association was not statistically significant. A further very low quality study (Inaloo et al., 2011) of 1600 children and adolescents reported no such association between the presence of migraine and coeliac disease (OR 1.00, 95% CI: 0.23 to 4.24).

4.1.5.5. Evidence for the relationship between dental enamel defects and coeliac disease in adults

One very low quality study (El-Hodhod et al., 2012) of 860 children and young adults reported a significant association between the presence of dental enamel defects and coeliac disease compared to a healthy age and gender matched population (OR 9.36, 95% CI: 9.36 to 52.39).

4.1.6. Evidence to recommendations

All four sub-questions relating to the evidence for the recognition of coeliac disease were presented in tandem and discussed together. Therefore, the linking evidence to recommendation information will be presented for all four components of this question at the end of this chapter.

4.1.7. Recommendations & research recommendations

All four sub-questions relating to the evidence for the recognition of coeliac disease were presented in tandem and discussed together. Therefore, the associated recommendations will be presented for all four components of this question at the end of this chapter.

4.2. Populations at increased risk of coeliac disease

4.2.1. Review question

What populations are at an increased risk of developing coeliac disease?

There is evidence to suggest that certain populations are at an increased risk of developing coeliac disease. It is necessary to identify these populations so that appropriate consideration for serological testing for coeliac disease can be made.

4.2.2. Methods

This review question concerns the identification of possible subgroups of people with coeliac disease. The aim of considering coexisting conditions was to examine whether people with certain conditions have a higher rate of coeliac disease than the general population while the estimation of familial risk is essential in a genetics-based condition such as coeliac disease. This focus on improving the identification of people with possible asymptomatic coeliac disease may also include active case finding in particular subgroups with a higher risk of coeliac disease. Included studies examined the presence of the following:

Coexisting diseases
Other factors (i.e. first-degree relatives)

The prevalence of coeliac disease in the populations studied was compared to a general population prevalence of 1.0%. An increased risk of coeliac disease is indicated when the confidence intervals around the prevalence in the population subgroup are all above 1.0% whereas the point estimate and confidence interval below 1.0% or crossing 1.0% indicates no increased risk compared to the general population.

4.2.3. Evidence review

A single systematic search was conducted for sections 4.1, 4.2, and 4.3 (see Appendix C) which identified 7230 references. This search was restricted to studies published from 2008 onwards to avoid duplicates of studies considered in the previous coeliac disease guideline (CG86). The references were screened on their titles and abstracts and full papers of 336 references were obtained and reviewed against the inclusion and exclusion criteria in the review protocol (see Appendix C).

Overall, 69 studies which examined the prevalence of coeliac disease in other conditions or first-degree relatives were included (see Appendix C). The remaining 267 studies were excluded. Reasons for exclusion are listed in Appendix F.

Description of included studies

Addison’s disease

A single study (Fichna et al., 2010) investigated 85 adults with autoimmune Addison’s disease. The age of study participants ranged from 18 to 82 years and 61 (71.8%) were female.

Arthritis

Three studies (Atzeni et al., 2008; Coacciloli et al., 2010; Francis et al., 2002) investigated a total of 222 adults with arthritis. The age of study participants ranged from 20 to 84 years and in total 147 (66.2%) were female. 195 (87.8%) had rheumatoid arthritis and 27 (12.2%) had psoriatic arthritis.

Juvenile arthritis

Three studies (George et al., 1996; Lepore et al., 1996; Robazzi et al., 2013) investigated a total of 224 children and young people with juvenile arthritis. The age of study participants ranged from 2 to 16 years and 148 (66.1%) were female.

Cardiomyopathy in children

A single study (De Menzes et al., 2012) investigated 56 children and young people with cardiomyopathy. The age of study participants ranged from 12 months to 18.8 years and 32 (57.1%) were female.

Cardiomyopathy

Three studies (Chicco et al., 2010; Frustaci et al., 2002; Vizzardi et al., 2008) investigated a total of 637 adults with cardiomyopathy. The age of study participants ranged from 28 to 92 years and 206 (32.3%) were female.

Down’s syndrome

Five studies (Bonamico et al., 2001; Cerqueira et al., 2010; Goldacre et al., 2004., Pavlovic et al., 2012; Wouters et al., 2009) investigated a total of 2999 children, young people and adults with Down’s syndrome. For children and young people the age ranged from 2 months to 18 years while for adults it ranged from 18 years to 59 years. Across all studies 732 (47.3%) were female, however gender was not specified in one study (Goldacre et al., 2004).

Epilepsy or seizures

Four studies (Cronin et al., 1998; Djuric et al., 2010; Peltola et al., 2009; Pratesi et al., 2003) investigated a total of 605 children, young people and adults with epilepsy or seizures. The age of study participants ranged from 12 months to 64 years and 305 (50.4%) were female.

Dyspepsia

A single study (Giangreco et al., 2008) investigated 726 children and adults with dyspepsia. The age of study participants ranged from 8 to 75 years and 44 (6.1%) were female.

Irritable bowel syndrome

Five studies (Cash et al., 2011; Cristofori et al., 2014; El-Salhy et al., 2011; Sanders et al., 2001; Sanders et al., 2003) investigated 3232 children, young people and adults with irritable bowel syndrome. The age of study participants ranged from 4 to 80 years but was not reported in 2 studies (Sanders et al., 2001; Sanders et al., 2003). 1264 (86.6%) were female; however this was not reported in two studies (Sanders et al., 2001; Sanders et al., 2003).

Other Gastrointestinal (GI) conditions

Five studies (Aziz et al., 2010; Casella et al., 2010; Leeds et al., 2007; Lynch et al., 1995; Simondi et al., 2010) investigated 2547 adults with ‘other GI conditions’. The age of study participants ranged from 18 to 80 years and, where reported, 1324 (52.4%) were female, however mean age and gender were not reported in one study (Lynch et al., 1995) and gender was not reported in Cristofori et al., 2014.

Liver disease

Nine studies (Bardella et al., 1997; Chatzicostas et al., 2002; Dickey et al., 1998; Drastich et al., 2012; Eapen et al., 2011; Gatselis et al., 2012; Germenis et al., 2005; Olsson et al., 1982; Thevenot et al., 2007) investigated 2955 adults with liver disease. The age of study participants ranged from 6 to 85 years was although age was not reported in 1 study (Olsson et al., 1982). Where reported, 1342 (46.3%) were female, however gender was not reported in 2 studies (Eapen et al., 2011 & Olsson et al., 1982)

Neurological disease

A single study (Ruggieri et al., 2008) investigated 300 children and young people with known neurological disorders. The age of study participants and gender were not reported.

Sarcoidosis

A single study (Papadopoulos et al., 1999) investigated 78 adults with sarcoidosis. The age of study participants ranged from 22 to 81 years and 34 (43.6%) were female.

Sjogren syndrome

A single study (Szodoray et al., 2004) investigated 111 adults with Sjogren syndrome. The age of study participants ranged from 28 to 77 years and gender was not reported.

Systemic sclerosis

A single study (Forbess et al., 2013) investigated 72 adults with systemic sclerosis. Mean age was 51years (SD = 13) and 66 (88%) were female.

Auto-immune thyroid disease

Three studies (Saatar et al., 2011; Sategna-Guidetti et al., 1998; Spadaccino et al., 2008) investigated 725 children and adults with autoimmune thyroid disease. The age of study participants ranged from 3.1 to 80 years. 612 (84.4%) were female.

Turner syndrome

Four studies (Bonamico et al., 2002; Dias et al., 2010; Frost et al., 2009; Mortensen et al., 2009) investigated 808 girls and women with Turner syndrome. Mean age ranged from 10 months to 61 years.

Type 1 diabetes

Twelve studies (Adlercreuttz et al., 2014; Barbato et al., 1998; Cev et al., 2010; Djurić et al., 2010; Galván et al., 2008; Kakleas et al., 2010; Leeds et al., 2011; Pham-Short et al., 2010; Picarelli et al., 2005; Salardi et al., 2008; Smith et al., 2000; Uibo et al., 2010) investigated 9014 children, young people and adults with type 1 diabetes. The age of study participants ranged from 12 months to 70 years and, where reported 3472 (49.5%) were female, however gender was not reported in one study (Salardi et al., 2008).

First-degree relatives

Nine studies (Almeida et al., 2008; Ascher et al., 1997; Biagi et al., 2009; da Silva Kotze et al., 2013; Esteve et al., 2006; Oliveira et al., 2012; Rubio-Tapia et al., 2008; Szaflarska-Szczepanik et al., 2001; Vaquero et al., 2014) investigated 3358 siblings or parents of people with biopsy-confirmed coeliac disease.

4.2.4. Health economic evidence

An economic literature search was not conducted for this question as an economic evaluation would not be the correct framework in which to generate useful evidence on the clinical conditions which can coexist with coeliac disease.

The populations identified within the clinical evidence review will be carried forward for consideration of the cost-effectiveness of obtaining a diagnosis of coeliac disease in these groups – see section 4.4.4.

4.2.5. Evidence statements

This review found that the following population subgroups had an increased risk of coeliac disease compared with the a background population prevalence of 1.0%

Autoimmune thyroid disease – pooled prevalence of 2.4% (95%CI 1.5 to 3.8%) low quality evidence from 3
Dyspepsia – prevalence of 2.1% (95%CI1.3 to 3.4%) low quality evidence from 1 study
Down’s syndrome - pooled prevalence of 3.2% (95%CI 1.3 to 7.4%) low quality evidence from 5 studies
Epilepsy or seizures – pooled prevalence of 3.6% (95%CI 1.9 to 6.7%) very low quality evidence from 4 studies
Sjogren syndrome – prevalence of 4.5% (95%CI 1.9 to 10.1%) low quality evidence
Turner syndrome – pooled prevalence of 5.5% (95%CI 4.1 to 7.4%) low quality evidence from 5 studies
Type 1 diabetes – pooled prevalence of 6.0% (95%CI 4.0 to 8.9%) low quality evidence from 12 studies; 3.3% (95%CI 2.4 to 4.6%) in 1 UK-based study
First-degree relatives – pooled prevalence of 8.2% (95%CI 4.6 to 14.3%) low quality evidence from 9 studies

This review found that the following population subgroups were at no increased risk of coeliac disease compared with a background population prevalence of 1.0%

Addison’s disease - 1.2% (95%CI 0.0 to 6.4%). very low quality evidence from 1 study
Arthritis - pooled prevalence of 3.0% (95%CI 0.8 to 11.0%) low quality evidence from 3
Juvenile arthritis - pooled prevalence of 2.3% (95%CI 0.9 to 5.3%) very low quality evidence from 3 studies
Cardiomyopathy - prevalence of 2.2% (95% CI 0.7% to 6.4%) very low quality evidence from a single study
Cardiomyopathy in children - prevalence of 1.8% (95% CI 0.3% to 9.5%) very low quality evidence from a single study
Irritable bowel syndrome - pooled prevalence of 1.8% (95%CI 0.7 to 4.7%) low quality evidence from 5 studies; 4.3% (95%CI 2.7 to 6.7) in 2 UK-based studies
Other gastrointestinal conditions - pooled prevalence of 2.9% (95%CI 0.5 to 16.6%) low quality evidence from 5 studies
Liver disease - pooled prevalence of 2.0% (95%CI 0.7 to 5.8%) low quality evidence from 9 studies
Neurological disease - prevalence of 1.1% (95%CI 0.5 to 2.3%) very low quality evidence from a single study
Sarcoidosis - prevalence of 0% low quality evidence from a single study
Systemic sclerosis - prevalence of 0% low quality evidence from a single study

4.2.6. Evidence to recommendations

4.2.7. Recommendations & research recommendations

4.3. Conditions associated with undiagnosed/untreated coeliac disease

4.3.1. Review question

What are the long-term consequences of undiagnosed or untreated coeliac disease?

It is estimated that only one fifth of those with coeliac disease are currently diagnosed. This indicates that up to four out of every five people with coeliac disease are currently untreated and at risk of serious long-term health complications. It is imperative to understand the nature of these long-term complications in order to understand the risks of untreated CD, particularly in populations who are deemed ‘at risk’ for this condition, as explored in section 4.2.

4.3.2. Methods

The aim of this question was both to identify information to be provided to people at the time of diagnoses (in the case of untreated coeliac disease) and to identify those long-term consequences of undiagnosed coeliac disease.

The GDG agreed to include studies which used either biopsy, according to Marsh 3 histological criteria, or positive serological test results to confirm the diagnosis of coeliac disease. The results for both are presented separately within this chapter. The GDG agreed to consider prevalence studies as well as studies estimating risk of having coeliac disease compared to a control group of age and gender matched participants.

The GDG however made a post-hoc decision to include only those studies where an estimation of the risk was presented. The GDG considered that this was easier evidence to interpret and would be more useful in clinical practice. The group also felt that have two different types of evidence for the same reviewing question could be confusing and potentially misleading.

A single systematic search was conducted for sections 4.1, 4.2, and 4.3 (see Appendix C) which identified 7230 references. This search was restricted to studies published from 2008 onwards to avoid duplicates of studies considered in the previous coeliac disease guideline (CG86). The references were screened on their titles and abstracts and full papers of 161 references were obtained and reviewed against the inclusion and exclusion criteria in the review protocol (see Appendix F).

Eleven studies (Canavan et al., 2011; Duerksen et al., 2010; Godfrey et al., 2010; Hogen-Esch et al., 2011; Jafri et al., 2008; Kumar et al., 2011, Leboff et al., 2013; Lohi et al., 2009; Sanchez et al., 2011; Silano et al., 2007; Zugna et al., 2013) were included (see Appendix C). The remaining 150 studies were excluded. Reasons for exclusion can be found in Appendix C.

Results are presented as adjusted hazard ratios, odds ratios, risk ratios or standardised incidence ratios.

4.3.3. Evidence review

Osteoporosis

One study (Jafri et al., 2008) investigated 83 children and adults with biopsy-confirmed coeliac disease. The age at diagnosis of coeliac disease range from 12 months to 84 years and 58 (70%) were female.

Four studies (Duerksen et al., 2010; Godfrey et al., 2010; Leboff et al., 2013; Sanchez et al., 2011) investigated 976 people with serology-confirmed coeliac disease. The age of participants ranged from 16 to 87.7 years and 869 (89.0%) were female.

Infertility

One study (Hogen-Esch et al., 2011) investigated 1038 male-female couples with infertility, of whom 10 individuals has unrecognised coeliac disease. The age of study participants ranged from 20 to 45 years.

No studies including biopsy-confirmed coeliac disease were identified.

Malignancy

One study (Silano et al., 2007) investigated 1968 people with biopsy-confirmed coeliac disease. The mean age at diagnosis of coeliac disease was 36.2 ± 13.8 years and 1485 (75.5%) were female.

Three studies (Canavan et al., 2011: Godfrey et al., 2010; Lohi et al., 2009) investigated 14503 people with serology-confirmed coeliac disease. The age of study participants ranged from 30 to 95 years and 8186 (56.4%) were female.

Increased mortality rate

One study (Zugna et al., 2013) investigated 16121 women with biopsy confirmed coeliac disease of whom 3202 (19.9%) were undiagnosed prior to giving birth. The study was concerned with the associated child mortality rate. The age categories used in the study ranged from 15 to 45 years.

No studies including serology-confirmed coeliac disease were identified.

4.3.4. Health economic evidence

4.3.5. Evidence statements

4.3.5.1. Biopsy-confirmed coeliac disease

Osteoporosis

Very low quality evidence from a single study reported the risk of osteoporosis, reported as any fracture, to be an adjusted hazard ratio of 2.0 (95% CI 1.0 to 3.9)

Very low quality evidence from a single study reported the risk of osteoporosis, reported as risk of peripheral fractures, to be an adjusted hazard ratio of 2.0 (95% CI 1.0 to 3.9)

Very low quality evidence from a single study reported the risk of osteoporosis, reported as risk of axial fractures, to be an adjusted hazard ratio of 1.7 (95% CI 0.7 to 4.2)

Very low quality evidence from a single study reported the risk of osteoporosis, reported as risk of osteoporotic fractures, to be an adjusted hazard ratio of 6.9 (95% CI 0.7 to 7.65)

Malignancy

Very low quality evidence from a single study reported the risk of malignancy, reported as non-Hodgkin’s lymphoma, Hodgkin’s lymphoma, small bowel, colon, oesophageal, melanoma, breast, stomach or other cancer to be a standardised incidence rate of 1.3 (95% CI 1.0 to 1.7)

Low quality evidence from a single study reported the risk of malignancy, reported as small bowel cancer to be a standardised incidence ratio of 25 (95% CI 8.5 to 51.4)

Low quality evidence from a single study reported the risk of malignancy, reported as non-Hodgkin’s lymphoma to be a standardised incidence ratio of 4.7 (95% CI 2.9 to 7.3)

Low quality evidence from a single study reported the risk of malignancy, reported as Hodgkin’s lymphoma to be a standardised incidence ratio of 10 (95% CI 2.7 to 25)

Low quality evidence from a single study reported the risk of malignancy, reported as stomach cancer to be a standardised incidence ratio of 3 (95% CI 1.3 to 4.9)

Very low quality evidence from a single study reported the risk of malignancy, reported as colon cancer to be a standardised incidence ratio of 1.1 (95% CI 0.7 to 1.6)

Mortality

Very low quality evidence from a single study reported the risk of increased child mortality (all cause), to be an adjusted hazard ratio of 1.1 (95%CI 0.9 to 1.3)

Very low quality evidence from a single study reported the risk of increased child mortality (non-accidental), to be an adjusted hazard ratio of 1.3 (95%CI 0.7 to 2.6)

4.3.5.2. Serology-confirmed coeliac disease

Osteoporosis

Low quality evidence from a single study reported the risk of osteoporosis to be an odds ratio of 2.6 (95% CI 1.3 to 5.1)

Low quality evidence from a single study reported the risk of osteoporosis, reported as fracture risk, to be hazard ratio of 1.5 (95% CI 1.1 to 2.1)

Low quality evidence from a single study reported the risk of osteoporosis, reported as T-score less than 2.5 to be a hazard ratio of 2.7 (95% CI 1.2 to 2.0)

Very low quality evidence from a single study reported the risk of osteoporosis, reported as, low bone mineral density (osteoporosis or osteopenia) to be an odds ratio of 1.0 (95% CI 0.1 to 95.8)

Malignancy

Very low quality evidence from a single study reported the risk of malignancy, reported as, coeliac disease related cancers, to be an odds ratio of 2.0 (95% CI 0.3 to 14.4)

Low quality evidence from a single study reported the risk of malignancy, reported as, lymphoproliferative cancers, to be an adjusted risk ratio of 5.9 (95% CI 1.4 to 25.0)

Very low quality evidence from a single study reported the risk of malignancy, reported as, breast cancer, to be an adjusted risk ratio of 0.7 (95% CI 0.1 to 5.1)

Very low quality evidence from a single study reported the risk of malignancy, reported as, all cancers, to be an adjusted risk ratio of 0.7 (95% CI 0.3 to 1.1)

Very low quality evidence from a single study reported the risk of malignancy, reported as mortality due to cancer, to be an adjusted risk ratio of 1.2 (95% CI 0.5 to 2.7)

Infertility

Very low quality evidence from a single study reported the risk of undiagnosed coeliac disease in those with infertility due to ovulation disorders to be an odds ratio of 5.4 (95% CI 0.9 to 32.3)

Very low quality evidence from a single study reported the risk of undiagnosed coeliac disease in those with male factor infertility to be an odds ratio of 5.4 (95% CI 0.9 to 32.3)

Very low quality evidence from a single study reported the risk of undiagnosed coeliac disease in infertile (any cause) women to be an odds ratio of 2.4 (95% CI 0.5 to 12.1)

Very low quality evidence from a single study reported the risk of undiagnosed coeliac disease in infertile (any cause) men to be an odds ratio of 0.9 (95% CI 0.2 to 4.1)

Low quality evidence from a single study reported the risk of undiagnosed coeliac disease in unexplained infertility in women to be an odds ratio of 4.5 (95% CI 1.4 to 19.2)

4.3.6. Evidence to recommendations

4.3.7. Recommendations & research recommendations

4.4. Active case-finding

4.4.1. Review question

Should active case-finding be implemented in people with coexisting conditions/subgroups that are associated with an increased risk of coeliac disease?

There are certain populations of people that have an increased risk of developing coeliac disease, as explored in section 4.2. Understanding the utility of case-finding in these populations will inform whether active case finding should be implemented in any of those populations. Case finding aims to increase diagnosis of coeliac disease in people who have the condition but are currently unaware and untreated in order to minimise the potential for the development of serious long-term health consequences of untreated CD, as outlined in section 4.3.

4.4.2. Methods

The aim of this review was to establish if patients with specific health conditions or specific subgroups with an increased risk of coeliac disease should be proactively investigated for coeliac disease.

4.4.3. Evidence review

A systematic search was conducted (see Appendix C) which identified 1483 references. References were screened on their titles and abstracts and full papers of 30 references were obtained and reviewed against the exclusion and inclusion criteria in the review protocol (see Appendix C).

All of these studies were excluded for reasons such as not being a primary study i.e. comment or letter to editor, or inappropriate study design i.e. following-up only serologically-positive individuals with a biopsy. .

In addition, the 1902 papers that were identified in the searches sections 5.1 and 5.2 (serological testing) were re-reviewed based on title and abstract in order to identify any studies that may meet the inclusion and exclusion criteria for the present question. No studies of relevance to the present review question were identified in this database.

4.4.4. Health economic evidence

4.4.4.1. Systematic review of published cost–utility analyses

An economic evaluations filter was applied to the search protocol for this review question with the aim of finding economic evaluations that explored the cost effectiveness of active case-finding for coeliac disease in at-risk subgroups.

The search identified 236 references. The references were screened on their titles and abstracts and 20 full-texts were ordered.

Four cost–utility analyses were found of relevance to the question: Mein & Ladabaum (2004) and Mohseninejad et al. (2013) explored testing people with irritable bowel syndrome (IBS) for coeliac disease; Swigonski et al. (2006) looked at case-finding in children with Down’s syndrome; and Dretzke et al. (2004) analysed children newly diagnosed with type 1 diabetes.

4.4.4.2. Original health economic analysis

An original cost–utility model was used to explore the benefits, harms and costs associated with serological investigation of people at increased risk of coeliac disease. A modified version of the model developed to analyse the serological investigation of people with symptoms suggestive of coeliac disease was used (see 5.2.4.2). In addition to the various testing strategies, an arm was simulated in which no testing was offered, in order to estimate the value of case-finding compared with none. The GDG prioritised 4 different populations in which to investigate this question: first-degree relatives of people with coeliac disease, people with irritable bowel syndrome, people with type 1 diabetes and people with autoimmune thyroid disease. This choice was based on the populations in which the GDG believed there was greatest current uncertainty and/or variation in practice.

Parameters that differed between these populations were prevalence of coeliac disease, baseline health-related quality of life and life expectancy. Prevalence estimates were drawn from the evidence synthesis conducted as part of the clinical review identifying populations at an increased risk of developing coeliac disease (see 4.2). On GDG advice, UK-specific data from this review were used, where they were available; if no UK-only studies were found for the population in question, the pooled value for all included studies was used. For first-degree relatives, type 1 diabetes and autoimmune thyroid disease, separate analyses were conducted for adults and children; for irritable bowel syndrome, adults only were considered, as the GDG advised that irritable bowel syndrome is a very uncommon diagnosis in children. Full details of the methods and results of the model are provided in Appendix G.

4.4.5. Evidence statements

No clinical evidence that met the inclusion and exclusion criteria for this question was found.

4.4.5.1. Health economic evidence statements

Evidence for the cost effectiveness of screening first-degree relatives of people with coeliac disease

An original, directly applicable cost–utility analysis with minor limitations estimated that case-finding in adult first-degree relatives of people with coeliac disease results in improved quality of life at increased cost, with an ICER of £14,000 per QALY gained. The ICER remained below £20,000 as long as it could be assumed that a gluten-free diet improves the health-related quality of life of people with subclinical coeliac disease by 1.24% or more (compared with a base-case estimate of 1.48%).

An original, directly applicable cost–utility analysis with minor limitations estimated that case-finding in child first-degree relatives of people with coeliac disease results in improved quality of life at increased cost, with an ICER of £18,800 per QALY gained. The ICER remained below £20,000 as long as it could be assumed that a gluten-free diet improves the health-related quality of life of people with subclinical coeliac disease by 1.36% or more (compared with a base-case estimate of 1.48%).

Evidence for the cost effectiveness of screening people with type 1 diabetes for coeliac disease

An original, directly applicable cost–utility analysis with minor limitations estimated that case-finding in adults with type 1 diabetes results in improved quality of life at increased cost, with an ICER of £17,100 per QALY gained. The ICER remained below £20,000 as long as it could be assumed that a gluten-free diet improves the health-related quality of life of people with subclinical coeliac disease by 1.50% or more (compared with a base-case estimate of 1.48%).

An original, directly applicable cost–utility analysis with minor limitations estimated that case-finding in children with type 1 diabetes results in improved quality of life at increased cost, with an ICER of £20,600 per QALY gained. The ICER fell below £20,000 if it could be assumed that a gluten-free diet improves the health-related quality of life of people with subclinical coeliac disease by 1.94% or more (compared with a base-case estimate of 1.48%).

A partially applicable health economic analysis with potentially serious limitations looking at testing for coeliac disease in children newly diagnosed with type 1 diabetes (Dretzke et al., 2004) found that screening with EMA is the most cost-effective option in this group of children.

Evidence for the cost effectiveness of screening people with autoimmune thyroid disease for coeliac disease

An original, directly applicable cost–utility analysis with minor limitations estimated that case-finding in adults with autoimmune thyroid disease results in improved quality of life at increased cost, with an ICER of £26,000 per QALY gained. The ICER fell below £20,000 if it could be assumed that a gluten-free diet improves the health-related quality of life of people with subclinical coeliac disease by 1.74% or more (compared with a base-case estimate of 1.48%).

An original, directly applicable cost–utility analysis with minor limitations estimated that case-finding in children with autoimmune thyroid disease results in improved quality of life at increased cost, with an ICER of £28,300 per QALY gained. The ICER fell below £20,000 if it could be assumed that a gluten-free diet improves the health-related quality of life of people with subclinical coeliac disease by 2.44% or more (compared with a base-case estimate of 1.48%).

Evidence for the cost effectiveness of screening people with irritable bowel syndrome for coeliac disease

An original, directly applicable cost–utility analysis with minor limitations estimated that case-finding in adults with irritable bowel syndrome results in improved quality of life at increased cost, with an ICER of £20,800 per QALY gained. The ICER fell below £20,000 if it could be assumed that a gluten-free diet improves the health-related quality of life of people with subclinical coeliac disease by 1.64% or more (compared with a base-case estimate of 1.48%).

A partially applicable health economic analysis with potentially serious limitations looking at testing for coeliac disease in people with IBS (Mohseninejad et al., 2013), found that screening is likely to be cost effective in people experiencing diarrhoea or mixed symptoms (diarrhoea and constipation) of IBS. Excluding the group of patients with symptoms of only constipation improves the cost effectiveness of screening.

A partially applicable health economic analysis with potentially serious limitations looking at testing for coeliac disease in people with symptoms consistent with an IBS diagnosis (Mein & Ladabaum, 2004), found that screening in this population is cost effective.

Evidence for the cost effectiveness of screening children with Down’s syndrome for coeliac disease

A partially applicable health economic analysis with potentially serious limitations looking at testing for coeliac disease as a way to prevent lymphoma in asymptomatic children with Down’s syndrome (Swigonski et al., 2006), found that quality of life does not improve and costs increase when compared with not screening this population.

4.4.6. Evidence to recommendations

Table

4.4.7. Recommendations & research recommendations

Offer serological testing for coeliac disease to:
- People with any of the following:
  - persistent unexplained abdominal or gastrointestinal symptoms
  - faltering growth
  - prolonged fatigue
  - unexpected weight loss
  - severe or persistent mouth ulcers
  - unexplained iron, vitamin B12 or folate deficiency
  - type 1 diabetes, at diagnosis
  - autoimmune thyroid disease, at diagnosis
  - irritable bowel syndrome (in adults)
- first-degree relatives of people with coeliac disease.
Consider serological testing for coeliac disease in people with any of the following:
- metabolic bone disorder (reduced bone mineral density or osteomalacia)
- unexplained neurological symptoms (particularly peripheral neuropathy or ataxia)
- unexplained subfertility or recurrent miscarriage
- persistently raised liver enzymes with unknown cause
- dental enamel defects
- Down’s syndrome
- Turner syndrome.
Advise people who have tested negative for coeliac disease, particularly first-degree relatives and people with type 1 diabetes, that:
- coeliac disease may present with a wide range of symptoms and
- they should consult their healthcare professional if any of the symptoms listed in recommendations 1 or 2 arise or persist.

Publication Details

Copyright

Publisher

National Institute for Health and Care Excellence (NICE), London

NLM Citation

Coeliac Disease: Recognition, Assessment and Management. London: National Institute for Health and Care Excellence (NICE); 2015 Sep. (NICE Guideline, No. 20.) 4, Evidence for the recognition of coeliac disease.

Relative value of different outcomes	Signs and symptoms The GDG recognised a lack of evidence for the signs and symptoms of coeliac disease (CD), and in particular the most commonly recognised presenting signs and symptoms such as gastrointestinal dysfunction, weight loss, and abdominal pain. The group discussed this and agreed that, because CD is such a well-established disorder in terms of recognition of the common features, there is no impetus to conduct research into this area, and therefore no evidence to support established clinical knowledge. The GDG further recognised that differentiating between symptoms in terms of those that should prompt clinicians to offer serological testing, and those where clinicians should consider serological testing, is further made difficult by the lack of supportive evidence to differentiate between these two classes of recommendations. Coexisting conditions and active case-finding The GDG raised the importance of increasing recognition of CD, which is widely underdiagnosed in the UK. Outlining which particular coexisting conditions have an increased risk of CD is of utmost importance in order to increase awareness for, and testing for, CD in these populations. This can be difficult due to an overlap or masking of CD-like symptoms with symptoms of coexisting conditions. The group noted that they would expect a gain in health-related quality of life after a diagnosis of CD was made in those with coexisting conditions; however no evidence was found for this outcome. Long-term complications The GDG felt that raising awareness of CD to increase diagnosis was of particular importance in order to minimise the likelihood of the development of serious long-term complications.
Trade-off between benefits and harms	Signs and symptoms of CD The GDG was clear about the importance of serological testing for CD in any person where a clinical suspicion has arisen. The group cited the estimate that 4 out of 5 people with CD are currently undiagnosed, and that it is of utmost importance to improve diagnoses of these individuals by increasing both clinical and community awareness of CD and the associated signs and symptoms. The GDG agreed that there were certain signs and symptoms and coexisting conditions that are sufficiently associated with CD that people with them should be offered serological testing, and developed recommendations to reflect this. The GDG further discussed the non-specific nature of many of the signs and symptoms and consequently added ‘unexplained’ and ‘chronic’ to the description of some signs and symptoms to ensure that people who may have CD are identified. Neurological symptoms were discussed in detail, as the group recognised that the literature to support suspicion of CD in this population was scarce. However, it was noted that a considerable number of individuals were detected by neurologists on the basis of recommendations in the previous guideline, which changed their practice substantially and subsequently led to a greater awareness of CD in patients with neurological symptoms. For this reason the GDG was convinced that serological testing should be considered in populations with neurological symptoms, especially ataxia or peripheral neuropathy, which have been reported in numerous case reports. The GDG agreed a list of further signs, symptoms and coexisting conditions for which they wanted to raise awareness of the link with coeliac disease. Therefore recommendations were developed that identified where offering serological testing for CD should be considered. The GDG also recognised that prolonged fatigue was a very common presenting feature of a myriad of disorders, both physical and psychiatric. However, members of the group cited research by Hin et al. (1999) which suggests that up to 3% of those who present with unexplained prolonged fatigue were positive for CD antibodies. The group also cited their vast clinical anecdotal experience in which many people who had previously thought of themselves as asymptomatic retrospectively recognised that they had been very tired for up to a decade before diagnosis was made. The importance of addressing the cause of prolonged fatigue was also raised as of high importance in paediatric patients, in whom fatigue is highly uncommon. Active case-finding The GDG emphasised that anyone who has symptoms suggestive of CD should be offered serological testing regardless of any coexisting conditions or characteristics. Therefore, the population of interest for the assessment of case-finding strategies should comprise people who are not currently experiencing such symptoms to a degree that leads them to seek advice from healthcare professionals. Following the conventions of the Oslo consensus statement on definitions for coeliac disease and related terms (Ludvigsson et al., 2012), the GDG preferred to refer to this group of people as experiencing ‘subclinical’ CD. This term is preferable to ‘asymptomatic’ disease, as it is clear that many people with undiagnosed CD have a history of symptoms that are retrospectively considered significant once a diagnosis has been established; moreover, it is common for people to report an improvement in such symptoms when they start a gluten-free diet (GFD). Therefore, people with subclinical CD should not be considered truly asymptomatic; instead, they are defined as people who experience ‘disease that is below the threshold of clinical detection without signs or symptoms sufficient to trigger CD testing in routine practice’ (Ludvigsson et al., 2012). First-degree relatives Current practice is to offer serological testing to first-degree relatives. The assumption that people do or do not have CD at the time of testing is incorrect. People may undergo seroconversion, which is problematic as a clinician may tell someone that they are not CD positive, but that person may develop CD at a later time. Ruling HLA DQ2/DQ8 out is important, as a clinician can then definitively conclude that if someone suspected of CD does not have HLA DQ2/DQ8, then they will never develop CD. While this could be very useful, it is pragmatically very difficult as a GP cannot request HLA DQ2/DQ8 testing as this needs to be requested by a specialist. Thus, patients would have to be referred to a specialist to request this test, which becomes expensive and time consuming, and therefore, in the opinion of the GDG, impractical. Type 1 diabetes The GDG raised the important notion that it is not sufficient to just test adults who present with gastrointestinal (GI) symptoms, as suggested in the current diabetes guideline. When people present at a diabetic clinic they are commonly only asked about diabetic features i.e. sugar, eyes, feet, etc., and GI symptoms are not discussed as part of a patient’s diabetic review so go unnoticed and therefore untested. It is estimated that 15–20% of people have GI symptoms, but people don’t realise that these may be relevant to their diabetes and so do not raise it with their diabetes consultant. The group felt strongly that it was very important to have a low threshold for testing people with diabetes to optimise dietary management of their diabetes and their potential CD-related symptoms. The GDG discussed current recommendations within the diabetes guideline relating to testing for CD when a low BMI is noted. The group discussed that weight loss and low BMI are a feature of CD and noted that, although weight loss can be a symptom of CD, the traditional view of a person with CD being underweight is no longer true and that people with diabetes may present underweight, at a normal weight or overweight. It is therefore important that low BMI should not be highlighted as the only circumstance in which suspicion of CD should be raised in someone with diabetes. The GDG also discussed testing for CD at the time of diagnosis of diabetes, and noted that this could be problematic in some circumstances. The group recognised that it may be too emotionally or cognitively difficult for the patient to take on the importance of each of their separate diagnoses. For these reasons, the GDG considered that it would be reasonable for a short delay (unlikely to be more than 6 months) between diagnosis of type 1 diabetes and testing for CD. The group chose not to complicate its recommendations with explicit discussion of this issue, as it believed that most clinicians would use their discretion in providing tests and information in an appropriate timeframe, and it did not want to detract from the importance of providing the test for everyone who has received a diagnosis of type 1 diabetes. Long-term complications The GDG felt that the available evidence highlighted the very serious nature of the potential long-term complications of undiagnosed CD. Osteoporosis was felt to be the most common potential long-term complication and the GDG felt that the evidence adequately reflected clinical experience. The GDG noted that, although there is an increased risk of malignancy with undiagnosed CD, the overall risk of developing specific cancers is low. The evidence for infertility was somewhat inconsistent. However, due to the serious emotional impact infertility has on a couple trying to conceive, the group still felt that it was important to raise awareness that CD could be contributing to this, and that clinicians should consider serological testing if other causes of infertility have been ruled out. Overall, the group felt strongly that serological testing is inexpensive and non-invasive, and that, if potentially very serious long-term complications could be avoided by having a diagnosis of CD made, the benefit of doing so far outweighs the potential detriment in having to follow a GFD. This trade-off was explored explicitly in original health economic modelling (see below).
Economic considerations	Active case-finding The original health economic analysis for this question was based on a modified version of the model developed to compare various serological testing strategies. Therefore, many of the considerations discussed in that question apply here (see 5.2.6). It was a potential weakness of the analysis that no evidence was found to estimate the diagnostic accuracy of different testing strategies in the populations of interest. Therefore, it was assumed that the sensitivity and specificity of the tests did not differ between populations, and data from the review of diagnostic accuracy in people presenting with symptoms suggestive of CD were used (see 5.1.3 and 5.2.3). In the original health economic model, the benefits of identifying people with subclinical CD are captured in 2 ways. Firstly, the quality of life of the proportion of people who follow advice to adopt a GFD will improve. Secondly, those people are subject to reduced incidence of long-term complications of CD, some of which have an impact on life expectancy. The GDG understood that, in all the populations simulated in the model, a reduction in long-term complications (with attendant improvement in life expectancy) was not, on its own, sufficient to counterbalance the costs and harms of testing (including serological assays and endoscopic biopsy in people who test positive). In contrast, the day-to-day quality of life benefit associated with a true-positive diagnosis only had to be small to make case-finding good value for money. The quality of life evidence used in the model’s base case was drawn from an Argentinian study in which quality of life was measured (using the SF-36) at the point of diagnosis and following 3 months’ treatment with a GFD. This suggested that people with subclinical CD who adopt a GFD experience quality of life that is, on average, approximately 1.5% better than those who continue to ingest gluten (Nachman et al. 2009). Although the study appears to have been well conducted, the sample of patients of interest to this model is very small; as a result, the estimate of effect is very uncertain. However, this uncertainty is appropriately propagated through the model, which presents a probabilistic synthesis of all parameters. The GDG expressed a clear view that it was appropriate to make a base-case assumption that adopting a GFD improves quality of life in people who were not complaining of symptoms at the time of diagnosis. Members of the group advised that, in their experience, many people who are diagnosed with subclinical CD report a history of symptoms that, while troublesome, had not led them to seek medical advice. Furthermore, the GDG reported that such people commonly report an improvement in such symptoms when starting a GFD. Finally, the fact that most people who have been diagnosed with subclinical disease elect to continue with a GFD is an indication that they are conscious of a perceptible improvement in quality of life. The GDG understood that a difference in quality of life of the magnitude used in the model’s base case to estimate the benefit of a GFD for people with subclinical CD is very small (1.5%). For comparison, the smallest effect that is detectable by the EQ-5D instrument and UK tariff (that is, the smallest change in quality of life that would result from an improvement in a single domain score) is equivalent to more than a 4% improvement in quality of life. Therefore, it was reasonable to assume that, if the quality of life of an average person with subclinical CD who adopts a GFD is improves by a degree that is perceptible to that person, a gain of at least 1.5% – and probably greater – on a quantitative measure could be expected. In this context, the base-case value should be seen as conservative. In all 4 populations simulated, the original health economic model suggested that case-finding in adults is likely to represent reasonable value for money. Base-case ICERs ranged between £14,000 per QALY gained (first-degree relatives) and £26,000 per QALY gained (autoimmune thyroid disease) for the best serological strategy compared with no testing. Case-finding was slightly more expensive in children than in adults, largely due to the increased costs associated with endoscopic biopsy in children (which usually requires anaesthesia). Nevertheless, case-finding resulted in improved quality of life, with ICERs ranging between £18,800 per QALY gained (first-degree relatives) and £28,300 per QALY gained (autoimmune thyroid disease). Although base-case ICERs exceeded £20,000 in type 1 diabetes (children only) and autoimmune thyroid disease (children and adults), the GDG felt these were likely to be somewhat underestimated, as the model only captured health gains that are associated with the diagnosis and management of CD. However, the group believed that, in both these conditions, correct identification of CD would also lead to superior management of the underlying condition, with associated improvement in quality of life. In the case of type 1 diabetes, the glycaemic control of people with subclinical CD is known to be improved by adopting a GFD. Additionally, dietary management is complex in people with both conditions, as each imposes its own requirements; in this context, the GDG believed it is critical for children to have access to appropriate dietetic support, so diagnosis of subclinical CD is very important. In the case of autoimmune thyroid disease, untreated coeliac enteropathy interferes with the absorption of oral medications that are critical to managing the condition. Correct identification of CD, therefore, should be associated with more stable and effective medication requirements, improving the person’s quality of life. In both these instances, the GDG felt that, although the additional benefits would be very hard to quantify without a complicated model of 2 concurrent disease processes, they were examples of ‘change in the quality of life [that] is inadequately captured’ in the analysis and, therefore, good reasons to recommend case-finding in populations that had base-case ICERs in the range £20–30,000. In all cases, results were very sensitive to the degree to which a GFD was assumed to improve the health-related quality of life of people with subclinical CD. However, the GDG felt confident that such benefits are observed in practice, so the group was happy to recommend case-finding, on the expectation that the true-positive identification of people with subclinical CD would lead to this kind of health gain. Although it is theoretically possible that different serological strategies might be optimal in different populations (according to expected prevalence of CD and other population-specific characteristics), little evidence was found to suggest that anything other than the strategies recommended in section 5.2 should be preferred. Therefore, it was not necessary to make separate recommendations about the tests that should be used in a case-finding context; it was sufficient to recommend that serological testing should be offered, and recommendations elsewhere in the guideline would be followed. One potential exception to this rule was that, in sensitivity analysis for child first-degree relatives of people with coeliac disease, some results suggested that it could theoretically be worth adding routine genotyping (HLA DQ2/DQ8 testing) to the diagnostic strategy. However, the GDG pointed out that, in practice, this would be of very limited value: if one family member is HLA DQ2/DQ8 positive (as the index case almost certainly would be), the chances of the rest of that family being HLA DQ2/DQ8 positive is very high. Therefore, the utility of doing that test in further family members is negligible. This shows that there are some areas in which population-specific diagnostic accuracy data might improve the accuracy of results. The original health economic model did not cover children with Down’s syndrome, as this population was not among the GDG’s top priorities for modelling. However, the GDG was presented with details of a published cost–utility analysis (Swigonski et al., 2006), which found that screening was not cost effective in this population. This analysis was confined to a single outcome of preventing lymphoma and the original health economic analysis conducted in other populations had shown that relatively little of the benefit of true-positive identification of CD could be ascribed to this outcome. Therefore, it was unsurprising that Swigonski et al. found insufficient benefit to justify the costs of case-finding. The GDG inferred that a fuller analysis, accounting for a wider range of benefits, would be likely to reach a different conclusion. However, the group did not feel that it had enough evidence to support an ‘offer’ recommendation, so concluded that case-finding should be considered in this population.
Quality of evidence	The group recognised that overall the quality of evidence available to answer this question was of a low quality. This was recognised to be a product of the lack of evidence available, the retrospective nature of the majority of studies, and the bias inherent in the way study participants were selected, how prevalence estimates were generated, the lack of precision in the presented estimates, and the lack of endoscopic intestinal biopsy to prove CD diagnosis in a great number of the studies available.
Other considerations	Irritable Bowel Syndrome (IBS) The GDG felt that it was important that those with a diagnosis of IBS should be tested for coeliac disease, as the two conditions have very similar phenotypic manifestations, notably in terms of gastrointestinal symptoms and abdominal pain. The group discussed that children are not routinely diagnosed with IBS, and are more likely to be labelled with ‘recurrent abdominal pain’ or ‘abdominal migraine’. IBS diagnosis is only typically given to adults with the same symptoms. Children may also have a diagnosis of inflammatory bowel disease. A child could hypothetically present to a number of clinicians and be given a number of different diagnoses for same symptoms because of this lack of consistency in characterising ‘IBS-like’ symptoms in the paediatric population. It was also raised as common for children to be diagnosed with IBS-like symptoms rather than a diagnosis of IBS. The group further raised the notion that, technically, children should be covered by recommendations on children or adults with recurrent GI symptoms, so whether this is labelled as IBS or not in children it is essentially irrelevant to their being investigated for CD. Repeat serological testing The group thought it highly important that both patients and healthcare professionals should be aware that people with risk factors for CD who test negative initially may remain at increased risk of developing CD in the future. The GDG was aware that some researchers have recommended routine periodic testing of people whose initial serological results are negative, especially those with type 1 diabetes. The original health economic model suggested that testing people with type 1 diabetes for CD at diagnosis could probably be considered to provide reasonable value for money (see above); however, the group was aware that this conclusion was relatively finely balanced, and small adjustments to the parameters of the model would produce different results. In particular, if prevalence of CD was any lower than estimated in the model’s base case, it would not be cost effective to offer case finding. The model was not designed to examine the cost effectiveness of periodic repeat testing. However, it could be inferred that the prevalence of CD among people who initially tested negative would be lower than in the incident type 1 diabetes cohort. Therefore, it is extremely unlikely that repeat testing would achieve health gains at a cost that would be considered an effective use of NHS resources. This result would arise partially because of the costs of repeat serology itself, but more particularly because of the costs incurred and quality of life forgone by performing endoscopic biopsies in people with positive serology in a context where those people were more likely to have false-positive findings owing to lower prevalence of CD (that is, the positive predictive value of all serological tests would be lower in a retesting setting). Nevertheless, the GDG were keen to emphasise that, if people with risk factors for CD who have previously been found to be serologically negative develop CD-like symptoms over time, there should be a low threshold for retesting for CD. The group were mindful of the evidence and experience (noted above) that people with subclinical CD frequently experience mild symptoms that do not lead them to seek medical advice. Therefore, the group recommended that people who initially test negative are advised to treat any future CD-like symptoms seriously, and not to hesitate to seek advice from their healthcare providers. For treating clinicians, it was emphasised that, if a person has risk factors for CD and even mild symptoms that are suggestive of CD, a historical negative serological test should not be used as a reason not to offer repeat serological testing.