Dr. Kate E Evans, Room P39, Department of Gastroenterology and Liver unit, Royal Hallamshire Hospital, Sheffield, South Yorkshire, S10 2JF, UK. (fax: 0114-271-2692; e-mail: email@example.com).
Abstract. Evans KE, Sanders DS (Royal Hallamshire Hospital, Sheffield, UK). What is the use of biopsy and antibodies in coeliac disease diagnosis? (Symposium). J Intern Med 2011; 269: 572–581.
The advent of highly sensitive and specific serological markers has led to some protagonists proposing that coeliac disease can be diagnosed without the need for a biopsy. However, this is an area of controversy. Lack of consensus about diagnostic degrees of histological change, paucity of symptoms, antibody-negative disease and immunodeficiency can make diagnosis difficult even with a biopsy. Conversely, an argument can be put forward for a ‘no biopsy’ approach based on the large number of patients with typical symptoms and positive serology who experience a diagnostic delay. In addition, endoscopy is not without discomfort. This article discusses the use of antibodies and duodenal biopsy within this context. Finally, we propose a pragmatic diagnostic algorithm for clinicians to use when investigating patients for coeliac disease.
To diagnose coeliac disease (CD), we must first define what we now consider CD to be. The first diagnostic criteria for coeliac disease (CD) were laid down in 1969 by the European Society of Paediatric Gastroenterology and Nutrition (ESPGAN) . Since this original definition, the evolution of endoscopy, development of accurate serological tests and large epidemiological studies have contributed to a growing knowledge base. Improved understanding of the wide variety of clinical manifestations of CD, the spectrum of histological change and recognition of associated serological markers and genetic links have all resulted in an evolving definition of CD.
The original ESPGAN criteria comprised three parts . First, there should be demonstration of structurally abnormal jejunal mucosa when taking a diet containing gluten. Secondly, there should be clear improvement of villous structure when taking a gluten-free diet (GFD). Thirdly, there should be deterioration of the mucosa during gluten challenge. This series of three biopsies reflects the perception of CD at that time; a permanent condition that started in childhood.
This concept was challenged in the 1980s with the development of simple, accessible and reliable serological tests and improved awareness of CD. These advances led to the recognition that the majority of patients have minor presenting symptoms [2, 3]. Concurrently, the concept of a ‘precoeliac’ state was introduced, and the terms latent and potential coeliac disease were first described.
The new millennium saw the first criteria for the diagnosis of CD in adults . The requirements for diagnosis were villous atrophy with crypt hyperplasia and intraepithelial lymphocytosis (IELs) whilst on a gluten-containing diet that normalizes on GFD. The finding of positive IgA tissue transglutaminase (tTG) or endomysial antibody (EMA) was deemed supportive, but not necessary. Likewise, a compatible human leucocyte antigen (HLA) type (DQ2 or 8) was considered supportive, but not essential.
Latent coeliac disease, potential coeliac disease and gluten sensitivity
The first descriptions of a precoeliac state were in patients with dermatitis herpetiformis and normal jejunal biopsies who developed typical enteropathy after a gluten challenge [5, 6]. The term latent CD was coined to describe a patient with a normal small bowel mucosa whilst on a gluten-containing diet who later develops CD (Fig. 1). Conversely, it may also describe a patient with biopsy-proven CD that later has a normal small bowel biopsy despite a normal gluten-containing diet. One recent study included 61 adults who had biopsy-proven CD in childhood . These patients were no longer adherent to the GFD. All were asymptomatic despite being on gluten-containing diet for upwards of 3 years. Of the 61, 13 had normal mucosa at repeat biopsy. Two of the thirteen with latent disease relapsed at subsequent follow-up. In this carefully selected group, it appears that up to 18% (11/61) may have true latency and in time develop immune tolerance to gluten.
Conversely, it has been shown that before the development of villous atrophy, patients may have symptoms such as abdominal pain, weight loss and diarrhoea or complications such as osteoporosis [8–13]. Furthermore, these may resolve with GFD [9–11, 14]. Individuals in this group, termed potential CD, may have a combination of features including normal duodenal biopsies or raised IELs, positive serology [EMA or significantly raised tissue transglutaminase (tTG)], the presence of raised gamma/delta T-lymphocytes or a positive rectal gluten challenge (the latter two features are considered research tools) [15–17]. More recently, the National Institute for Health consensus group used the term latent CD to describe individuals with positive serology and normal biopsy . This has been adopted by some researchers.
Gluten sensitivity is a term that has been used interchangeably with potential CD and has become an area of increasing research interest. Gluten sensitivity may be defined as a condition of some morphological, immunological or functional disorder that responds to gluten exclusion . The concept would include gluten-sensitive diarrhoea, immunological response to gluten in family members of CD patients and positive serology in the absence of mucosal injury .
Assessing the usefulness of a test
The usefulness of a diagnostic test, which is its ability to identify an individual with disease or exclude one without, is usually described in terms of sensitivity, specificity, positive predictive value (PPV) and negative predictive values (NPV). The sensitivity of a test is defined as the proportion of people with disease who have a positive test. A highly sensitive test is useful for ruling out a disease if a person tests negative . The specificity of a test is the proportion of people without disease who will have a negative result. A highly specific test is useful for ruling a disease in. Sensitivity and specificity are important measures of the diagnostic accuracy of a test but cannot be used to predict disease probability in an individual. Because both are defined on the basis of people with or without disease, they cannot be used to estimate the probability of disease in a patient before they have a diagnosis. Instead, PPV and NPV describe an individual’s likelihood of disease. The PPV of a test is defined as the proportion of people with a positive test that actually has the disease. Conversely, the NPV is the proportion of people with a negative test result who do not have disease. Importantly, both PPV and NPV vary according to disease prevalence, and so predictive values for one population cannot be applied to another with a different prevalence of disease. It is essential to understand these limitations when considering serological testing for CD.
Endomysial antibody and tissue transglutaminase
EMA testing is highly accurate with a sensitivity and specificity of 95% or more in patients with overt villous atrophy [21–25]. However, it is subjective, labour intensive, and the substrates (monkey oesophagus, umbilicus) are limited . tTG assays are generally cheaper than EMA and more reliable [21, 27]. One weakness of the tTG test is that the accuracy of the assay varies between manufacturers . The best assays have a higher sensitivity than EMA and a comparable specificity, both around 98% [27–32]. The cohort studies that comprise the majority of the evidence for the performance of each test are of high quality but too heterogenous to provide pooled data . Instead, ranges for sensitivity and specificity are given (see Table 1) [24, 27–31, 33–45].
Table 1. Sensitivity and specificity of coeliac serologies
Sensitivity range (%)
Specificity range (%)
DGP, deamidated gliadin peptides; EMA, endomysial antibody; IgA, immunoglobulin A; POCT, point of care testing; tTG, tissue transglutaminase.
Although EMA and tTG appear to be sensitive and specific, these observations are based on carefully selected high CD prevalence populations. In lower population prevalence (for example 1%), the PPV of the test falls. When CD prevalence falls below approximately 35%, the PPV of tTG and EMA falls from 90 to 100% to 80% or less . In a low prevalence population as seen in screening, the specificity of the test has to be near perfect for the PPV to remain above 90% .
Our group performed serological testing and concurrent duodenal biopsies on 2000 consecutive adults attending for gastroscopy . We identified 77 new cases of CD (seven antibody negative). In this referral population, tTG had a sensitivity and specificity of 91%, an NPV of 99%, but a PPV of just 28%. EMA had a PPV of 71%, and an NPV of 99%. This study highlights the poorer performance of the tests in this heterogenous group which, perhaps, more accurately reflects typical clinical practice.
The sensitivity of the serological tests also falls when histological grades less than Marsh 3 are considered. In these circumstances, the sensitivity falls well below 90% [30, 47, 48]. This is a clinical problem that is difficult to evaluate as most studies have excluded patients without villous atrophy .
Importantly, the relationship between adherence to GFD, mucosal recovery and serology is not linear. It is possible to have normalization of serology with persisting mucosal atrophy . There is no reliable, objective marker for adherence and mucosal recovery [50, 51]. Likewise, a subjective clinical response to GFD is not predictive of mucosal recovery . Indeed, despite a clinical improvement on GFD, complete mucosal recovery is rare . The best way to assess adherence is by skilled dietary history .
Because the main EMA and tTG tests are IgA based, they are prone to error in conditions associated with abnormal levels of IgA. For example, IgA deficiency is associated with CD and is a cause of false-negative serological results [55–59]. The near perfect NPV of EMA does not take into account IgA-deficient patients. Conversely, false-positive tTG may also occur and is associated with conditions of raised IgA such as chronic liver disease and monoclonal gammopathy [60, 61]. One option in the presence of IgA deficiency is to use IgG EMA or IgG tTG antibody tests .
Deamidated Gliadin peptides (DGPs)
Although IgA and IgG gliadins have been superseded by EMA and tTG antibody, more recently, a new generation of assays detecting the presence of deamidated synthetic peptides of gliadin (a-DGP) has shown high diagnostic performance equivalent to conventional tests [36–38, 63] A recent meta-analysis does not suggest any diagnostic advantage is conferred by the use of deamidated gliadin peptides  In addition, these novel serological tests are currently more expensive than the established options.
Point of care testing
Point of care testing (finger prick testing) has an apparent advantage as these tests can be applied in the office/outpatient or endoscopy setting and potentially provide the clinician with an immediate answer about the risk of having CD. Initial results are promising, but the reported sensitivities and specificities are still less than conventional serology [39–45, 64]. Further work is required in this area particularly testing in low-risk populations.
No biopsy strategy
A no biopsy strategy for CD will inevitably be centred round serological testing. One group has evaluated such a strategy, but using a different cut-off level than that recommended by the manufacturer . Hill et al. performed a retrospective comparison of the tTG levels and biopsy findings of 112 patients with new diagnosis CD (7 Marsh Grade 2, 105 Marsh Grade 3). A tTG level 10× upper limit of normal (ULN) gave a PPV of 100%. Using a higher threshold for a positive result increases specificity but at the expense of sensitivity. There may be a role for this approach in patients who are unwilling or unable to have a gastroscopy and duodenal biopsy.
Coeliac disease is associated with the DQ2/DQ8 allelles in the HLA class II complex [66, 67]. HLA typing may be useful in excluding coeliac disease. A negative test for HLA DQ2 and DQ8 has a high negative predictive value which may have a role in those with borderline histology or serology, or in those unwilling to give up a self-imposed GFD [68–70].
A positive family history is another indication where HLA may be useful. One study tested 441 first-degree relatives for tTG, IgA and HLA typing . Those who were serology negative were followed up after 2–3 years. Of the 40 with positive serology, 38 carried DQ2 or DQ8. Three individuals who were initially antibody negative developed positive serology on long-term follow-up. All were DQ2 positive. This suggests that those who are HLA compatible for CD may need continued follow-up, whereas those who are HLA DQ2 and DQ8 negative could be discharged.
HLA status may also be used to risk stratify first-degree relatives . Recent work has suggested that based on a gene dose effect, clinicians may quantify the risk of developing CD to be low risk or high risk, depending on the presence or absence of homozygosity [72–74].
HLA typing is more expensive than conventional serology but less than endoscopy . In one analysis, when the pretest probability of CD is less than 7% and the specificity of tTG <99.2, HLA screening was less costly than proceeding direct to endoscopy . As the pretest probability of CD increases, the number of individuals with positive serology rises, and the added cost of endoscopy should be weighed against the consequences of a false-positive diagnosis. In relatives, HLA screening may reduce gastroscopy workload and cost .
Evolving controversies related to duodenal biopsy
Historically, a small bowel biopsy was obtained using a suction biopsy capsule. With the advent of fibreoptic endoscopy, investigators were able to demonstrate that endoscopic duodenal biopsy was comparable to the suction capsule (Crosby capsule) in terms of its ability to detect villous atrophy [77, 78]. Approaches using multiple biopsies have demonstrated the patchy nature of CD in children and adults [70, 79, 80]. As a result, guidelines recommend multiple biopsies to minimize sampling error .
Conventional understanding is that villous damage is most severe proximally because this is where the gluten load is greatest . It might be expected therefore that a duodenal bulb biopsy is necessary to detect the most severe lesion. However, Brunner’s glands and the presence of gastric metaplasia were thought to interfere with histological interpretation of the bulb. Villi can be naturally shorter, blunted or absent in these areas [82, 83]. The recent advance that we and others have reported is that the optimal strategy for diagnosing CD can only achieve 100% sensitivity by always incorporating a duodenal bulb biopsy [46, 80, 84–86]. Our study included 56 patients with positive tTG or EMA . Biopsies were taken from the bulb, proximal and distal duodenum. Of the 56, 53 had villous atrophy present in at least one biopsy. Of the 53, 10 had patchy villous atrophy. In all patients with villous atrophy, it was detected by taking three biopsies – always including one from the bulb (sensitivity 100%, 95% CI 93–100). The most severe degree of villous atrophy was only detected using a five-biopsy regime. Only 3/56 (5.4%) had identical degrees of histological change at all nine sites.
Other causes of villous atrophy are rare. In one study of 2000 biopsies from the second part of the duodenum, only three patients had villous atrophy attributable to causes other than coeliac disease (two Helicobacter pylori, one Crohn’s disease) . Villous atrophy may be reported erroneously because of poor specimen orientation, but other causes of villous atrophy occur including, for example, peptic inflammation, Giardia, Helicobacter infection or Crohn’s disease [46, 82, 87].
However stringent a diagnostic algorithm is, the nature of medicine ensures that there will always be diagnostic dilemmas, or grey cases. Do those without villous atrophy have the same risks as those with? The evidence is mixed; they may have increased mortality from certain conditions such as ischaemic heart disease, but not others such as lymphoma . Do they need to be on a GFD? The answers are not known.
Among the untreated CD, 10–20% are EMA negative [88–91]. One study reported 22 IgA competent patients with EMA-negative disease . Three had small bowel lymphoma. The patients were older and had more abdominal symptoms suggesting more advanced disease. Of note, though antibodies were not measurable in the serum, they were deposited and detectable in the small bowel mucosa. Contrary to this, others report that antibody-negative disease is associated with milder histological lesions [30, 93].
Selective IgA deficiency (as previously discussed) is also a cause of antibody-negative disease. Finally, other possible causes of antibody-negative coeliac disease may be a self-imposed GFD prior to testing or concomitant use of immunosuppressant drugs.
Marsh grade 1 and 2 (potential coeliac disease)
Scott and Losowsky were the first to describe patients with suspected malabsorption who had small bowel biopsies that were not flat and did not meet the criteria for CD. All responded clinically and histologically to a GFD. More recently, Kurppa et al. randomized 23 EMA-positive patients with Marsh 1 or 2 changes and compatible HLA to normal (n = 10) or GFD (n = 13) . The patients that were randomized to a gluten-containing diet had progression of mucosal damage and persistence of seropositivity. Patients that were randomized to a GFD showed histological, serological and symptomatic improvement. This observation is supported by other investigators in a nonrandomized study . However, caution is needed because lesser histological changes are not uncommon and may not always be related to CD [96, 97]. One prospective study of 100 patients with Marsh 1 changes found that 16% had CD. Other causes included drugs, infection and immune dysregulation . Under these circumstances, a gluten challenge may also be another option [97, 98]. Likewise, response to GFD does not always indicate CD. One study reported the investigational findings of 112 patients on GFD without confirmatory serology or biopsy. Of the 112 patients, 51 were subsequently diagnosed with CD (based on a positive EMA and villous atrophy on biopsy). For the patients that had a negative EMA and normal biopsy, 75% reported that their symptoms had previously improved when they had been placed on a GFD; furthermore, 54% had recurrence of symptoms on reintroduction. This symptomatic response rate was not different from those with confirmed CD .
Capsule endoscopy (CE) allows visual inspection of the bowel without intubation and may be more acceptable to a patient than endoscopy. CE is able to detect villous atrophy and markers of CD such as reduced folds, mosaic pattern, visible vessels and scalloping [100–103]. In Marsh grade 3 disease, CE has a high sensitivity of 70–87.5% and specificity 90.9–100% [104–107]. Capsule endoscopy is less sensitive for Marsh grade 1 and 2 lesions [104–106]. A further limitation is that it is possible to have villous atrophy at biopsy and a normal CE [105, 106].
Augmented diagnostic clinical algorithm
When creating a diagnostic algorithm, it is useful to look at weaknesses in current practice. In a pragmatic study, Biagi et al. reported 614 patients referred for nonresponsive CD . In 434 (70%), there was unequivocal villous atrophy and positive serology. Reasons for doubting the diagnosis of CD in the remainder (n = 180/614) included lack of biopsy (n = 95), mild lesions (n = 63), negative serology (n = 61) and lack of response to GFD (n = 31). CD was ultimately excluded in 84–119 patients (13.6–19.4%). This was based on normal histology, normal IgA and negative serology. Alternative diagnoses were found in 48/84 including irritable bowel syndrome, food allergy, lactose malabsorption and infectious diarrhoea. The investigators had anticipated diagnostic difficulty because of poor quality biopsies, minimal intestinal lesions and antibody-negative disease. However, in practice, difficulties arose primarily through lack of biopsy or antibody testing.
Several investigators have shown that the combination of serological tests, duodenal biopsy and incorporating key symptoms or history (chronic diarrhoea, weight loss and family history) may result in a 100% detection strategy for CD. Using this approach, we and others have shown that patients can be stratified into high- and low-risk groups [46, 109].
When considering an algorithm, economic costs must be considered. With particular respect to CD, the relatively poor PPV of tTG must be balanced against the low sensitivity of EMA and high cost of endoscopy. Some authors have suggested a 2-step approach [49, 110, 111]. We performed an economic analysis of coeliac diagnosis in an unselected population attending for gastroscopy . A total of 2000 consecutive patients had gastroscopy with duodenal biopsy and tandem tTG, and EMA. Of the 2000, 77 were diagnosed with CD. The overall prevalence of IgA deficiency was 14/2000 (0.7%) and one of these had CD. Six patients had EMA and tTg-negative CD, one had EMA positive only. Of the 2000, 245 were tTG positive. If this were used alone to determine biopsy, seven cases of CD would be missed, but this high sensitivity (90.9%) is offset by the high cost of endoscopy. If we took a 2-step approach where a positive tTG is followed by EMA and we then only performed a duodenal biopsy if the patient is EMA positive, then we would have missed four cases of CD, but the number of endoscopies undertaken would fall from 245 to 92, thus making the 2-step option cheaper. Dorn et al. included HLA typing in their potential diagnostic algorithms . Where pretest probability was low, it was cost effective to biopsy all tTG-positive individuals. However, as the pretest probability increased, so did the number of serology-positive patients and hence the number of gastroscopies. Costs could be reduced by HLA typing and only performing biopsy in those with compatible HLA.
We believe that an augmented diagnostic clinical algorithm is reflective of real practice, pragmatic and cost effective (Fig. 2).
Making the diagnosis of CD is now more complex than ever before (Fig. 2). It is the belief of the authors that a ‘cast iron’ diagnosis is essential at the outset. This should be based on a positive antibody (EMA or tTG) in the presence of villous atrophy. We have provided evidence to show that the serological tests cannot entirely replace histology (Table 1). Furthermore, patients without CD may respond to a GFD . Subjectively, we believe that patients are reassured by having a histological diagnosis. Would we accept a diagnosis of lung cancer based on a chest x-ray? The confirmatory histology leaves no ambiguity and may even improve adherence rates to the GFD. Baseline histology may allow objective comparison for both those patients seeking to be reassured by histological remission or for those with persisting symptoms. In the United Kingdom, some general practitioners may refuse to prescribe the GFD in the absence of clear evidence of coeliac disease. Finally, the importance of an absolute diagnosis does not just involve the index case but also our approach as clinicians to first-degree relatives. For those patients who fall into the grey or equivocal zone then referral to a gastroenterology, consultant is essential and within this context we must also ensure that these individuals are not committed to a GFD prior to further investigation.
Conflict of interest
KEE – no conflict of interest to declare. DSS – declares that he received Biocard kits without charge and is currently involved in an evaluation study. DSS is an associate medical adviser for Coeliac UK (National Medical Charity), an honorary reader in gastroenterology at the University of Sheffield, and chairman of the small bowel and nutrition committee of the British Society of Gastroenterology. These are honorary posts with no financial benefits.
There are no acknowledgements that need to be made on the script. There has been no funding.