Definition and prevalence

Coeliac disease is the manifestation of an immune-mediated hypersensitivity reaction towards gluten in genetically-predisposed individuals. Treatment is with a lifelong gluten-free diet (GFD); indeed, part of the definition of coeliac disease is that it improves clinically and histologically on gluten withdrawal. However, a proportion of patients do not improve on a GFD, and it is these patients who will be the focus of the present review.

Non-responsive coeliac disease (NRCD) has historically been described as failure to respond to a GFD and can be manifest as primary NRCD (the individual never responds to a GFD) or secondary NRCD (a subsequent recurrence of symptoms after 1 year⁽Reference Ryan and Kelleher¹⁾, although a recent study has suggested that an alternative timescale for secondary NRCD could be 6 months⁽Reference Leffler, Dennis and Hyett²⁾). However, these definitions are all arbitrary. Furthermore, some studies have suggested that NRCD may not just be a failure to respond to a GFD but also the presence of persisting histological changes⁽Reference Ryan and Kelleher^1, Reference Abdulkarim, Burgart and See³⁾.

Historical studies estimate the prevalence of NRCD in the general population at 7–30%⁽Reference O'Mahoney, Howdle and Losowsky^4–Reference Pink and Creamer⁶⁾. However, these estimates were based on small cohorts of patients. Contemporary clinical studies or patient-support-group surveys from the USA and UK have suggested that NRCD occurs in about 17–30%⁽Reference Fine, Meyer and Lee^7–Reference Green, Stavropoulos and Panagi⁹⁾. The majority of cases of NRCD are likely to be related to either inadvertent ingestion of tiny but clinically significant amounts of gluten or non-adherence⁽Reference Leffler, Dennis and Hyett^2, Reference Abdulkarim, Burgart and See^3, Reference Fine, Meyer and Lee⁷⁾.

When determining whether a patient has NRCD it is important to be guided by symptoms rather than the histology alone. Clinical improvement may occur within days on a GFD but histological improvement continues for ⩽2 years. The distal small bowel improves most rapidly and proximal biopsies may not reflect this improvement⁽Reference Ryan and Kelleher^1, Reference Dickson, Streutker and Chetty^10, Reference Grefte, Bouman and Grond¹¹⁾. It has been demonstrated that after ≥1 year on a GFD less than one-quarter of patients with coeliac disease have normal endoscopic appearance or histology⁽Reference Lee, Winson and Memeo¹²⁾. In the thirty-nine patients who were investigated normal histology was found after a mean of 8·5 (range 1–45) years in 21% (n 8) of patients, with partial villous atrophy in 69% (n 27) and total villous atrophy in 10% (n 4). More recently, persisting abnormalities have been demonstrated in 82·5% of adult patients with coeliac disease⁽Reference Bardella, Velio and Cesana¹³⁾. After a median of 2 (range 1–23) years on a GFD 17·5% of the 114 adult patients studied were found to have normal histology (Marsh grade 0; for details of the Marsh classification of lesions, see Marsh⁽Reference Marsh¹⁴⁾), with 20·2% having raised intraepithelial lymphocytes (IEL; Marsh grade 1) and 30·7% and 31·6% having partial villous atrophy (Marsh 3a) and total villous atrophy (Marsh 3b) respectively. All patients were reported to have responded clinically to the GFD. It should be noted that the importance of raised IEL (in this context) without other histological change is still a matter of great debate.

Thus, it is recommended that classifying a patient as NRCD should be based primarily on persisting symptoms.

Does this patient really have coeliac disease?

In the patient with NRCD it is important to challenge the initial diagnosis of coeliac disease; in a report on one series of fifty-five patients referred for NRCD the original diagnosis of coeliac disease was disproved in six⁽Reference Abdulkarim, Burgart and See³⁾. All previous diagnostic investigations including the serology and the original and subsequent duodenal biopsies should be reviewed. Biopsies should be well orientated and interpreted by a specialist gastrointestinal pathologist⁽Reference Abdulkarim, Burgart and See³⁾.

If the diagnosis of coeliac disease is uncertain then supportive evidence should be sought in the form of typical human leucocyte antigen status (DQ2 or DQ8), a positive family history of coeliac disease (with an attributable risk of approximately 10% for first-degree members), functional hyposplenism and the detection of IgA or IgG gliadin antibodies (historically), a raised endomysial antibody or tissue transglutaminase antibody⁽Reference O'Mahoney, Howdle and Losowsky⁴⁾.

Other causes of malabsorption and/or villous atrophy that should be ruled out are shown in Table 1⁽Reference Abdulkarim, Burgart and See³⁾.

Table 1. Causes of malabsorption and/or villous atrophy⁽Reference Abdulkarim, Burgart and See³⁾

Continued exposure to gluten?

The most common reason for NRCD is continued gluten exposure. This cause can be substantiated by a composite approach: (1) the patient describing symptoms; (2) a dietetic assessment for adherence; (3) persisting antibody titres (either tissue transglutaminase antibody or endomysial antibody), particularly if there is a baseline titre for comparison; (4) the persistence of histological changes on duodenal biopsy. In the group of fifty-five patients described earlier twenty-five of the fifty-nine with coeliac disease were identified as having gluten contamination responsible for their ongoing clinical and histological signs⁽Reference Abdulkarim, Burgart and See³⁾. Inadvertent ingestion was found to usually occur in the form of commercially-packaged cereals derived from maize or rice that had contained malted barley. By recognising these sources as the reason for inadvertent gluten exposure it was shown that all patients improved symptomatically⁽Reference Abdulkarim, Burgart and See³⁾.

Adherence to a GFD is estimated to be approximately 50–70%⁽Reference O'Leary, Wieneke and Healy^15–Reference Hauser, Stallmach and Caspary¹⁷⁾. Whilst it is true that patients overestimate their adherence, factors such as cost, poor availability and poor palatability of gluten-free foods decrease compliance. In addition, socialising and eating out may lead to inadvertent exposure to gluten⁽Reference Karajeh, Hurlstone and Patel¹⁸⁾. Factors associated with improved compliance include understanding the GFD, membership of a coeliac disease advocacy group, attendance at a follow-up clinic and the perceived ability to maintain the diet in the course of everyday-life obstacles⁽Reference Bardella, Molteni and Prampolini¹⁹⁾.

What are the other possible causes of persisting symptoms?

Having established the diagnosis of coeliac disease and adherence to a GFD, the next step in the patient with NRCD is to exclude other causes of symptoms.

This process can be subdivided into diagnoses that may be associated with coeliac disease, including microscopic colitis, pancreatic insufficiency, irritable bowel syndrome, small bowel bacterial overgrowth, thyroid dysfunction and secondary intolerances (as a result of mucosal surface damage, e.g. lactose or fructose intolerance)⁽Reference Leffler, Dennis and Hyett^2, Reference Fine, Meyer and Lee⁷⁾. Less-common alternative diagnoses that are not associated with the condition but may occur in patients with coeliac disease include dietary allergies, other protein-losing enteropathies, complete malrotation of the gut, anal sphincter dysfunction and Whipple's disease⁽Reference Leffler, Dennis and Hyett^2, Reference Abdulkarim, Burgart and See^3, Reference Fine, Meyer and Lee⁷⁾.

Microscopic colitis is more common in patients with coeliac disease than in healthy controls; however, the recognition of this entity is only possible if random colonic biopsies are taken⁽Reference Leeds, Horoldt and Sidhu^20, Reference Williams, Kaplan and Makhija²¹⁾.

Pancreatic abnormalities have been noted historically in patients with coeliac disease⁽Reference Rabsztyn, Green and Berti^22–Reference Patel, Johlin and Murray²⁴⁾. More recently, it has been demonstrated that approximately one-third of those patients with coeliac disease who have chronic diarrhoea have low faecal pancreatic elastase levels. In addition, pancreatic supplementation confers some benefit, with a reduction in stool frequency⁽Reference Green, Stavropoulos and Panagi⁹⁾.

Irritable bowel syndrome can mimic the symptoms of coeliac disease⁽Reference O'Leary, Wieneke and Buckley^25, Reference Midhagen and Hallert²⁶⁾. A report on 150 patients with coeliac disease, thirty of whom fulfilled the Rome criteria (diagnostic criteria used to make a positive diagnosis of functional bowel disorders, such as irritable bowel syndrome, based on clinical symptoms⁽Reference Longstreth, Thompson and Chey²⁷⁾), shows that patients with coeliac disease who have irritable bowel syndrome-type symptoms have a markedly lower quality of life than those without irritable bowel syndrome⁽Reference O'Leary, Wieneke and Buckley²⁵⁾. In addition, the quality of life of these individuals partly improves with improved adherence to a GFD. It was hypothesised that mucosal inflammation in coeliac disease may have a sensitising effect.

In a group of fifteen patients with coeliac disease who were all described as showing histological improvement on a GFD but remaining symptomatic, it was reported that ten were showing small intestinal bacterial overgrowth (assessed by a lactulose breath test). When these patients were treated with antibiotics (rifaximin) they were found after 1 month to be symptom free⁽Reference Tursi, Brandimarte and Giorgetti²⁸⁾.

Adult autoimmune enteropathy is now recognised as a rare cause of villous atrophy. Patients have severe enteropathy unresponsive to any exclusion diet and are positive for enterocyte antibodies⁽Reference Corazza, Biagi and Volta^29, Reference Akram, Murray and Pardi³⁰⁾.

The clinical work up

Based on these data an initial work up, as shown in Fig. 1, is recommended. Routine blood tests, including haematinics (i.e. Fe, folic acid and vitamin B₁₂, which are necessary for Hb and erythrocyte production), should be undertaken. Repeat serological testing of tissue transglutaminase antibody and endomysial antibody, as well as human leucocyte antigen DQ typing, provide supportive evidence of coeliac disease. Blood tests including inflammatory markers and thyroid function tests are done to give supportive evidence of alternative diagnoses. Repeat gastroscopy with biopsy and small bowel aspirate should be performed. Colonoscopy with biopsies is done to exclude microscopic colitis. Faecal pancreatic elastase allows assessment of exocrine pancreatic function. Stools should be examined for giardia and other parasites. If these tests are normal, more unusual causes such as lactose or fructose intolerance or protein-losing enteropathy should be sought. In complicated cases with normal repeat duodenal biopsies faecal α-1-antitrypsin can be a useful marker to confirm that the patient has a protein-losing enteropathy.

Fig. 1. Diagnostic algorithm for non-responsive coeliac disease (NRCD). HLA, human leucocyte antigen; FHx, family history; SBBO, small bowel bacterial overgrowth; PLE, protein-losing enteropathy; RCD, refractory coeliac disease.

Don't forget patients who are super-sensitive to gluten!

It is important to remember that even if these tests are normal and the patient still has persisting symptoms the individual could be super-sensitive to gluten. The typical gluten content of a GFD is 1 mg/d⁽Reference Catassi, Rossini and Ratsch^31, Reference Ciclitira, Cerio and Fagg³²⁾; most patients with coeliac disease are able to tolerate ⩽10 mg/d without adverse clinical or histological signs⁽Reference Akobeng and Thomas³³⁾. As a practical guide, a 30 g slice of wheat bread typically contains 4·8 g gluten and the average diet contains 13 g gluten/d⁽Reference Stern, Ciclitira and van Eckert^34, Reference Ciclitira, Ellis and Lundin³⁵⁾. In the UK naturally-occurring foods that contain <20 mg gluten/kg are considered ‘gluten free’ or ‘naturally gluten-free’. Processed foods containing <100 mg gluten/kg are also considered gluten-free but the current exact labelling requirements are in the process of being clarified. Thus, patients could still have symptoms whilst adhering carefully to a GFD. Evidence of this outcome would be expected to be seen with the composite assessment made initially, but the use of a wheat-free diet and GFD may provide symptomatic benefit. However, this strategy is even more restrictive!

Pathogenesis

The normal small intestinal IEL population comprises approximately 80–85% CD8+T-cell receptor (TCR) αβ cells and 15% CD8+TCRγδ cells⁽Reference Rust, Kooy and Pena³⁶⁾. IEL are derived from oligoclonal expansion of a relatively small number of αβ T-cell clones. IEL are thought to recognise bacterial proteins, preserve epithelial integrity and mediate antigenic tolerance. In uncomplicated coeliac disease IEL express CD3+ and CD8+ (T suppressor–cytotoxic phenotype) and there is an increase in γδ T-cells⁽Reference Rust, Kooy and Pena^36, Reference Inagaki-Ohara, Chinen and Matsuzaki³⁷⁾.

Coeliac disease may be considered to be refractory (RCD) when symptoms persist (primary) or recur (secondary) despite the adherence to a strict GFD and when other causes have been excluded⁽Reference Ryan and Kelleher¹⁾. RCD is subdivided into types I and II with a phenotypically normal or aberrant intraepithelial T-cell population respectively⁽Reference Inagaki-Ohara, Chinen and Matsuzaki³⁷⁾. This type differentiation can be determined by using PCR analysis for the TCR gene rearrangement. RCD I is thought to yield a polyclonal product and in RCD II the gene rearrangement is clonal. Some authors have described a cut-off level in percentage terms as a means of differentiating RCD I from RCD II. Patients with RCD I have <10% aberrant (clonal) T-cells on duodenal biopsy⁽Reference Inagaki-Ohara, Chinen and Matsuzaki³⁷⁾.

This subdivision into RCD I and RCD II according to the percentage of aberrant T-cells has allowed investigators to infer differences in prognosis. RCD I is more likely to respond to immunosuppression, whereas RCD II seems largely resistant to treatment and transition to enteropathy-associated T-cell lymphoma (EATL) is common. In a group of ninety-three patients with RCD forty-three were reported to have RCD I, with none having coeliac-related mortality at 5 years. In the group with RCD II (n 50) the overall 5-year survival was reported to be 58%, falling to 8% in the twenty-six (52%) who developed EATL⁽Reference Cellier, Delabesse and Helmer³⁸⁾.

It is thought that T-cells bearing TCRγδ play a protective role in intestinal inflammation⁽Reference Ashton-Key, Diss and Pan³⁹⁾. In RCD II there is an immunophenotypically-aberrant clonal intraepithelial T-cell population with monoclonal arrangement of the TCRγ gene⁽Reference Verbeek, von Blomberg and Scholten^40, Reference Meijer, Mulder and Goerres⁴¹⁾. The presence of these abnormal T-cells is correlated with lymphoma development⁽Reference Love, Elmes, Golden, McNicholl, McCarthy and Fottrell⁴²⁾. Interestingly, in EATL the non-lymphomatous mucosa (adjacent mucosa) contains an identical monoclonal T-cell to the EATL, i.e. this condition could be cryptic lymphoma. The same observation has been made in ulcerated and non-ulcerated mucosa in ulcerative jejunitis. These observations could suggest that that RCD is a spectrum of disease ranging from the presence of monoclonal T-cells to ulcerative jejunitis and finally EATL. However, patients can present de novo at any point along this path without progressing through these stages.

A limited number of factors are predictive of RCD. In one series of patients weight loss was found to be predictive of RCD, with an OR of 31·1, and female patients were shown to have a decreased risk of RCD⁽Reference Leffler, Dennis and Hyett²⁾. Patients who are homozygous for human leucocyte antigen DQ2 appear more likely to develop secondary EATL or de novo EATL⁽Reference Al-Toma, Verbeek and Hadithi⁴³⁾. Typically, RCD presents with severe malnutrition, malabsorption, hypoalbuminaemia and weight loss⁽Reference Meijer, Mulder and Goerres⁴¹⁾. Ulcerative jejunitis can present similarly or as an emergency with perforation, obstructing stricture or bleeding. Lymphoma may present with local signs of abdominal pain and diarrhoea, or more systemically with weight loss, fever, sweats and skin rashes.

The clinical approach

A diagnostic approach to RCD is shown in Table 2⁽Reference Mulder, Wahab and Moshaver⁴⁴⁾. The initial steps are to record the BMI and then exclude NRCD as outlined earlier. Coeliac serology, human leucocyte antigen typing, routine blood tests, haematinics, Ca, Mg, Zn and Cu should all be checked. Deficiencies in Zn and Cu have been identified as possibly contributing to RCD and there have been case reports of successful treatment with supplementation⁽Reference Love, Elmes, Golden, McNicholl, McCarthy and Fottrell⁴²⁾. Duodenal biopsy should be assessed for T helper:suppressor cells, which is one way of initially differentiating patients with RCD from those who do not have RCD. It is worth specifically requesting the pathologist to assess the biopsy and comment on the ratio; a preponderance of T-helper cells suggests that RCD is more likely and a preponderance of T-suppressor–cytotoxic cells is consistent with normal coeliac disease and perhaps continued exposure to gluten. Specialist centres, including that of the authors, now offer assessment for aberrant clonal T-cell proliferation.

Table 2. Initial work up for refractory coeliac disease (RCD)Footnote *⁽Reference Mulder, Wahab and Moshaver⁴⁴⁾

IEL, intraepithelial lymphocytes; CT, computerised tomography; DEXA, dual-energy X-ray absorptiometry.

* Should follow the initial investigational pathway shown in Fig. 1 for patients in whom RCD is suspected.

Following the initial work up delineated in Table 2 further exclusion or inclusion of lymphoma may require an opinion from an ear, nose and throat specialist⁽Reference Askling, Linet and Gridley^45, Reference Catassi, Bearzi and Holmes⁴⁶⁾, bone marrow aspiration, ultrasound scan of the neck, full-thickness small bowel biopsy and laparotomy with intra-operative enteroscopy. The initial small bowel investigations are generally available to most clinicians but it is worth considering several modalities.

Newer techniques proving valuable in the assessment of RCD include positron emission tomography scanning, wireless capsule endoscopy and double-balloon enteroscopy. There are no large trials published to date, but in a small series of cases it has been shown that positron emission tomography has a high sensitivity for EATL and picks up some lesions missed on computerised tomography scanning⁽Reference Hadithi, Mallant and Oudejans⁴⁷⁾. However, false positive positron emission tomography results were also recorded in patients with RCD in the absence of EATL, which demonstrates the need to obtain histology. When wireless capsule endoscopy was performed in forty-seven patients with coeliac disease and a clinical suspicion of small intestinal malignancy, 87% were reported to have findings of coeliac disease with 60% having unexpected findings that included ulceration, cancer, polyps, stricture, submucosal mass and intussusception⁽Reference Culliford, Daly and Diamond⁴⁸⁾. Furthermore, in fourteen patients with RCD who underwent wireless capsule endoscopy EATL was found in two patients with RCD II, one case of which had not been detected by computerised tomography or MRI⁽Reference Daum, Wahnscaffe and Glasenapp⁴⁹⁾. Double-balloon enteroscopy can be used to obtain histology from lesions identified on MRI. In a group of twenty-one patients with RCD referred for double-balloon enteroscopy EATL was found in five patients, with another two having ulcerative jejunitis. EATL was excluded by double-balloon enteroscopy in four patients for whom there had been suggestive radiographic findings⁽Reference Hadithi, Al-Toma and Oudejans⁵⁰⁾.

Based on these small studies it is now believed that in the presence of histopathological findings consistent with RCD II EATL should be aggressively sought. The justification for this approach is that historically this condition has been difficult to detect and the diagnosis has important prognostic implications for the patient.

Therapeutic options

The evidence for treatment of RCD is limited and amounts to little more than case series. Azathioprine and prednisolone have shown some benefit in RCD I. In nineteen patients with RCD treated with azathioprine and prednisolone eight of ten patients with RCD I were reported to have responded. Conversely, in those with RCD II six of eight patients developed EATL and seven patients died⁽Reference Goerres, Meijer and Wahab⁵¹⁾.

More recently, budesonide has also been used in this clinical scenario with reasonable effect⁽Reference Brar, Lee and Lewis⁵²⁾. There are case reports of the successful use of cyclosporine⁽Reference Longstreth^53–Reference Wahab, Crusius and Meijer⁵⁵⁾. A single study assessing the use of IL-10 does not demonstrate benefit⁽Reference Mulder, Wahab and Meijer⁵⁶⁾. There is anecdotal use of total parenteral nutrition and albumin infusions but insufficient evidence to support their wider use.

Novel approaches to the treatment of RCD include anti-TNF⁽Reference Gillett, Arnott and Mcintyre^57, Reference Turner, Moorghen and Probert⁵⁸⁾, cladribine⁽Reference Al-Toma, Goerres and Meijer⁵⁹⁾ and autologous stem-cell transplantation⁽Reference Al-Toma, Visser and van Roessel^60, Reference Al-Toma, Verbeek and Visser⁶¹⁾. None has yet demonstrated clear benefit.