In a small percentage (2–5%) of adult-onset coeliac disease patients, refractoriness to a gluten-free diet or pre- and malignant complications develop.¹ Coeliac disease may be regarded as refractory (RCD) when symptoms persist or recur after a former good response despite strict adherence to a gluten-free diet.^1–3 We define RCD as persisting villous atrophy with crypt hyperplasia and increased intraepithelial T lymphocytes (IELs) in spite of a strict gluten-free diet for more than 12 months or when severe symptoms necessitate intervention independent of the duration of the diet.^2,4 Immunologically, two types of RCD are recognised depending on the presence or absence of aberrant IELs in the small-bowel mucosa. When normal expression of T-cell surface markers occurs (RCD I), the prognosis is less dismal than when an aberrant IEL population is present (RCD II).^2,4,5 Patients with RCD II are known to be at a greater risk of developing malignancy, particularly enteropathy-associated T-cell lymphoma (EATL).^2,5–7

There is now strong molecular and immunophenotypic evidence showing that a monoclonal neoplastic T-cell population may emerge from IELs in RCD. Clonal expansion of this T-cell population eventually leads to frank EATL. The genesis and expansion of these aberrant T cells involve both inappropriate immune responses to gluten and acquisition of genetic abnormalities. Although the IELs in RCD II can be neoplastic, they are not cytologically abnormal and do not form tumour masses that differentiate these patients from EATL patients, in addition to the absence of radiological and bone marrow evidence of lymphoma.^2,8,9,10

RCD II is usually resistant to any known therapy that has thus far been tested, including azathioprine/prednisone, cyclosporine and interleukin (IL)-10.^2,11–16 The response to cladribine (2-Chlorodeoxyadenosine) therapy is less than optimal.¹⁷ EATL has a poor outcome with current therapies, with 1- and 5-year survival rates in the range of 31–39% and 11–20%, respectively.^18–20

So far, no systematic analysis of the survival of this group of patients has been reported. The aim of this study is to provide insight into the course of RCD and EATL, by reporting on one of the largest cohorts of patients with complicated coeliac disease in a single centre. We have retrospectively compared the survival in four groups of patients: RCD-I, RCD II total, secondary EATL and de novo EATL.

PATIENTS AND METHODS

Patients

We performed a retrospective analysis, providing long-term follow-up data on four categories of patients with complicated forms of coeliac disease in a tertiary referral centre for coeliac disease. From 1992 to 2005, 43 patients with RCD I (12 males and 31 females; mean age at diagnosis of RCD: 49 years, range 23–86 years), 50 patients with RCD II (19 males, 31 females; mean age at diagnosis: 59 years, range 47–88 years)—of whom, 26 patients with secondary EATL (11 males, 15 females; mean age at diagnosis of EATL: 61.5 years, range 52–79 years) and 13 patients with de novo EATL (11 males, 2 females; mean age at diagnosis: 64.3, range 56–72 years) were studied. A small group of patients with RCD have been excluded from the analysis; they were treated with cyclosporine or IL-10.^14,15

The baseline characteristics of the patients according to the groups are shown in table 1. The patients with RCD I and II were followed for evidence of transition to a more severe state (ie, the transition from RCD I to RCD II and/or EATL, and from RCD II to EATL) over a mean period of 5 years (range, 2–14 years).

Diagnostic criteria

The diagnostic criteria of the different groups are summarised in table 3.

In RCD patients, the presence of EATL has been confidently excluded using radiological and endoscopic methods: small bowel follow-through, computed tomography (CT) scanning,²¹ whole-body positron emission tomography (PET),²² upper gastrointestinal endoscopy, video capsule endoscopy (VCE) and/or double-balloon enteroscopy (DBE)²³, as well as trephine bone-marrow biopsies. Those patients diagnosed in 2003 or earlier have negative small bowel follow-through and CT scans, whereas those diagnosed after 2003 have (in addition to negative small bowel follow-through and CT scans) negative PET, VCE and/or DBE.

Evaluation

Clinical, laboratory (haematology, biochemistry and serology), endoscopic and histological examination of the small intestine was performed at regular intervals (3–6 months). Clinically, patients were followed up at the outpatient clinic at regular intervals and their adherence to the diet advised was checked by a dietician. Particular attention was paid to clinical response and adverse effects of therapies.

Antiendomysium antibodies (EMA) and anti-tissue transglutaminase antibodies (anti-tTG) were tested at diagnosis and at follow-up. In all patients, HLA-DQ2/8 typing²⁴ and flow cytometric immunophenotyping of IELs²⁵ were performed.

Endoscopy was performed as indicated, using upper gastrointestinal endoscopy, VCE and/or DBE with small-intestinal biopsies.

CT scan, PET scan, magnetic resonance (MR) enteroclysis and dual-energy X-absorptiometry were performed as indicated. The techniques for VCE, DBE and MR enteroclysis have been available in our centre since the beginning of 2003.

Statistical analysis

Kaplan–Meier survival curves were constructed using SSPS software (SPSS Inc. Chicago, Illinois, USA) for comparison between the groups. Chi square test, odds ratio (OR), log rank and logistic regression tests were used to assess the statistical significance between variables. A p value ⩽0.05 was considered statistically significant.

RESULTS

Table 1 shows the baseline demographic characteristics according to the disease group. Regarding gender distribution in the studied groups, there was no difference between the RCD I and RCD II groups. In contrast, the de novo EATL group showed a statistically significant increase in the male:female ratio compared with the RCD I (p<0.001) and secondary EATL group (p<0.025).

Patients with RCD I were younger than RCD II and EATL patients, but no transition was documented from RCD I to RCD II or EATL during the period of follow-up.

The HLA-DQ2 genotype was present in 80% of RCD I patients, 92% of total RCD II patients, 96% of secondary EATL patients and 92% of de novo EATL patients. HLA-DQ2 homozygosity was significantly higher in secondary EATL and de novo EATL compared with RCD I patients: OR = 3.6 (confidence interval (CI) 2.64–6.33) and OR = 2.06 (CI 1.88–4.78), respectively.²⁶

Serologically, all patients with positivity for EMA and/or anti-tTG at the time of diagnosis of coeliac disease reverted to being negative after the gluten-free diet, confirming their strict adherence to diet (in addition to clinical assessment by a dietitian). Those with de novo EATL had negative serology at the time of diagnosis.

Concerning the mean percentage of aberrant T cells at the time of diagnosis of the RCD and/or EATL, the difference between RCD I (3%) versus RCD II (60%) and between secondary EATL (68%) versus de novo EATL (9%) was statistically significant (p<0.0001).

Ulcerative jejuntitis was not seen in any patient with RCD I. In the RCD II group, 5 patients of those treated with cladribine had ulcerative jejunitis and 4 of these who subsequently received autologous stem-cell transplantation (ASCT) had persistent histological recovery with disappearance of ulcerations.²⁷ None of these patients have so far developed EATL. Table 2 shows the treatments received and summarises protocols. Table 3 summarizes the diagnostic criteria of the different disease categories. From all EATL patients, 23 (59%) were treated with chemotherapy (cyclophosphamide, doxorubicine, vincristine and prednisone) (CHOP), whereas the other 16 patients were not eligible for chemotherapy due to a bad general condition at the time of diagnosis. Fourteen (36%) patients underwent partial resection of the small intestine. Laparotomy was needed in 3 patients to establish a diagnosis and 11 were operated for complications (5 for perforations and 6 for obstructive symptoms). Seven patients were treated both by chemotherapy and resection. Eight patients (61.5%) of the de novo EATL group underwent partial small-intestine resection compared with only 6 (23%) patients of the secondary EATL group (p = 0.025). Three other RCD II patients had surgery due to ulcerative jejunitis with perforations. Four patients with EATL were treated with high-dose chemotherapy followed by ASCT (3 with de novo EATL and one with secondary EATL) and 3 of them died within a few months thereafter.²⁰

Table 4 shows the causes of death according to patients’ categories. In the RCD I group, only 3 patients died during follow-up, all of them from unrelated illnesses. In the RCD II group, 26 patients (52%) developed EATL within 4–6 years after the diagnosis of RCD II had been made, and 23 RCD II patients (46%) died after developing EATL. Four (8%) patients died due to progressive malabsorption with emaciation; one patient developed extensive multifocal squamous-cell carcinoma of the skin (>15 lesions). One patient died due to progressive neurocoeliac disease 8 months after ASCT.²⁷ Nine of 13 (69%) patients with de novo EATL and 23 of 26 (88.4%) patients with secondary EATL died within months of diagnosis.

Figure 1 (A and B) shows the Kaplan–Meier curve of survival according to the disease group. The 5-year survival was 96% in RCD I versus 58% in RCD II (total) (p = 0.001). On the other hand, the 2-year survival in the de novo EATL group was 20% versus 15% in the EATL (RCD II) group (p = 0.63). Interestingly, the most significant drop in survival in these groups was observed in the first 2 years after diagnosis. The longest survival so far in the de novo EATL is 26 months. As this is not a randomised study, it is not possible to make definitive conclusions about the success of different treatments. However, in the prednisone-alone group, the 5-year survival was 25%; in the prednisone and azathioprine group, it was 36% (p = 0.43); and in the cladribine (2-CDA) group, it was 22% (43% at 36 months); p = 0.97. With respect to EATL development, there was no statistically significant difference between the groups.

Engraftment occurred in all RCD II patients who received ASCT. Neither major non-haematological toxicity nor transplantation-related mortality was observed. The mean follow-up duration was 16 months (range: 8–31 months). Within 3–4 months of ASCT, all patients had normalisation of stools frequency, disappearance of abdominal pain and improvement in biochemical markers. In addition, improvement of the body mass index, serum albumin, endoscopical findings and histology was documented. One patient with pre-existent neurocoeliac disease mimicking multiple sclerosis developed progression and died 8 months after transplantation. The details of the protocol, inclusion and detailed results are described in our recent article.²⁷

DISCUSSION

The development of a refractory state in patients with coeliac disease may herald the start of a serious phase in the evolution of the disease state—particularly in RCD patients with aberrant T cells (RCD II).^1,2 Particular attention should be paid to detect non-compliance or inadvertent gluten ingestion. Persistent positive serology may point to this latter scenario.^28–30

Regarding gender distribution in the studied groups, there is a female:male predominance in the RCD I, RCD II total and secondary EATL groups, whereas there is a reversed ratio with more males than females in the de novo EATL group. The difference is not significant between the RCD I and RCD II groups (p = 1.05). However, it is highly significant in the de novo EATL versus RCD I (p<0.001) groups and versus the secondary EATL group (p<0.025). Other studies also reported a predominance for EATL in males, with a peak in the sixth decade of life;^7,30,31 however, the majority of patients with uncomplicated coeliac disease are female.^32,33

HLA-DQ2 homozygosity is significantly higher in the secondary EATL and de novo EATL groups compared with the RCD I group: OR = 3.6 (CI 2.64–6.33) and OR = 2.06 (CI 1.88–4.78), respectively. As we reported earlier, there is an association between DQ2 homozygosity and complicated forms of coeliac disease.²⁶

A combination of prednisone and azathioprine is usually sufficient to treat RCD I patients.^11,16 None of our RCD I patients progressed to RCD II during follow-up. Cellier et al.² also reported 3 RCD patients without aberrant T cells, who made a complete recovery with steroid therapy plus a gluten-free diet. This underscores the value of performing T-cell flow cytometry in these patients, as the absence of aberrant T cells at diagnosis of the refractory state seems to indicate a favourable prognosis, and conventional treatment with prednisone with or without azathioprine is usually sufficient. In view of the poor prognosis of EATL, the question has always been: Can the diagnosis of lymphoma be made earlier to allow more effective treatment, thereby improving the prognosis? Despite state-of the-art technologies used, we can only assume that lymphoma has been excluded in these RCD II patients. The question remains as to whether a “subclinical” lymphoma was actually present and/or its development can be accelerated by available therapies.³⁴

Different therapies have been evaluated, but there is no effective therapy available yet for these RCD II patients.^2,11–17 Reports claiming good response are difficult to interpret due to an absence of clear distinction between RCD I and RCD II in these series. Cladribine therapy might be promising in stabilising a patient’s condition, and it improves the performance status and the histological picture, as seen in 58% of our group. However, it does not prevent EATL.¹⁷ Nine patients (39%) of the 23 treated patients died from EATL. High-dose chemotherapy, followed by ASCT after stabilisation with cladribine, might be an alternative approach in these pre-lymphoma patients. Our experience with the first 7 patients is encouraging, but it remains to be proven if development of EATL can be delayed or prevented.²⁷

Factors that seem to be associated with a high risk for EATL development in coeliac disease are old-age presentation, male sex, ulcerative jejunitis, presence of aberrant T cells and DQ2 homozygosity.^26,35,36 However, we have seen the disappearance of ulcerative lesions after cladribine therapy and/or ASCT in five patients, and none of these patients has developed EATL so far.^17,27 Thus, in the case of early intervention in the high-risk RCD II group—specifically, ulcerative jejunitis—EATL development might be prevented.

Patients with EATL can present in two different clinical patterns. There are patients with well-established coeliac disease who deteriorate because of the development of RCD II and eventually develop secondary EATL. In the other group, patients develop EATL without a preceding history of complicated coeliac disease; these patients often present with perforation or obstruction (primary or de novo EATL). The aberrant T cells in primary EATL patients appear to be largely confined to the tumour mass and cannot be found in such high percentages diffusely throughout the small intestine as in secondary EATL patients (p<0.0001), possibly suggesting a different pathogenesis pathway.

Nine of 13 (69%) of the de novo EATL patients and 23 of 26 of the secondary EATL patients (88.4%) died despite therapy. The 2-year survival in the de novo EATL group is 20% versus 15% in the EATL (RCD II) group (p = 0.63). Interestingly, 61.5% of the de novo EATL group had undergone resection compared with 23% of the secondary EATL group (p = 0.025). Howdle et al.⁷ reported a laparotomy rate of 73% in lymphomas associated with coeliac disease. It may be necessary to resort to laparotomy when lymphoma is suspected and when the diagnosis can not be established with less invasive methods. Overall, surgery was needed in 3 patients to establish a diagnosis and 11 had surgery for complications (5 for perforations and 6 for obstructive symptoms). Surgery, radiotherapy and chemotherapy may be used depending on the stage and clinical condition. EATL is often disseminated at diagnosis and has almost always a dismal outcome. However, if EATL is confined to part of the small intestine and if the affected segment (or segments) can be resected, the prognosis might be reasonable; some patients survive more than 5 years.^37,38 Debulking by surgery might be mandatory; however, prospective studies are lacking. Three of 4 EATL patients who received high-dose chemotherapy and ASCT died within months after transplantation.²⁰

Although the studied groups of patients seem to be rather heterogeneous and not entirely exclusive, we think that we have provided here a detailed description of the prognosis and response to currently available therapeutic options in the whole spectrum of complicated forms of coeliac disease—ranging from RCD I (with its relatively benign course) to RCD II (which has a definite pre-lymphoma potential to the frankly malignant EATL). Furthermore, this group of patients is one of the largest cohorts reported from a single centre dealing with coeliac disease and its complications. The most important papers dealing with RCD described series ranging from 1 to 16 RCD II patients.^{2,8,25,39–44}

In conclusion, an extensive evaluation and aggressive targeted management might be helpful in dealing with complicated forms of coeliac disease. Studies are needed to define more precisely the cut-off point between acceptably normal and pathologically increased percentages of aberrant T cells. Multicentre cooperation and studies are required to further increase the understanding of RCD in general.