Complete recovery of intestinal mucosa occurs very rarely in adult coeliac patients despite adherence to gluten-free diet


Prof. A. Lanzini, Gastroenterology Unit – Medicine 1, Piazza Spedali Civili 1. 25123 Brescia, Italy.


Background  Expected benefits of gluten-free diet (GFD) in coeliac patients include healing of small intestinal mucosa, but it remains unclear to what extent this benefit is achieved in adults.

Aim  To assess factors affecting histological outcome of GFD in a large cohort of adult coeliac patients.

Methods  We extracted information on 465 consecutive coeliac patients studied before and during GFD.

Results  Duodenal biopsies at diagnosis were classified as Marsh I in 11, II in 25 and III in 429 cases. After a median 16 months GFD, 38 (8%) patients had histological ‘normalization’, 300 (65%) had ‘remission’ with persistent intraepithelial lymphocytosis, 121(26%) had ‘no change’ and 6 (1%) had ‘deterioration’. Coeliac disease related serology was negative in 83% of patients with Marsh III lesion during GFD. Male gender and adherence to GFD were independently associated with histological ‘normalization’ and ‘remission’. Persistence of intraepithelial lymphocytosis was not associated with human lymphocyte antigen gene dose or with Helicobacter pylori infection.

Conclusions  Complete normalization of duodenal lesions is exceptionally rare in adult coeliac patients despite adherence to GFD, symptoms disappearance and negative CD related serology. Control biopsies are mandatory to identify lack of response to gluten-free diet.


Coeliac disease (CD) is a small-intestinal inflammatory disease with heterogeneous clinical presentation triggered by ingestion of gluten proteins of wheat, rye and barley in genetically predisposed individuals carrying the human lymphocyte antigen (HLA)-DQ2 or -DQ81. The only treatment of CD is lifelong adherence to a gluten-free diet (GFD) that is expected to improve not only symptoms and nutritional status but also to prevent long-term complications including malignant diseases.1 Clinical manifestations and complications of CD are thought to be pathogenetically related to ongoing duodenal mucosal inflammation,2 suggesting that healing of duodenal mucosa as achieved by GFD is a key factor to prevent complications.

Evidence in the literature on the extent of healing of duodenal mucosa during GFD is conflicting, with reported persistence of mucosal abnormalities ranging between 4%3 and 55%,4 a wide range that is likely to reflect differences among studies3–11 in sample size, mixed paediatric and adult populations and variable duration of GFD prior to control biopsies. Although poor adherence to GFD12 is the most obvious explanation for persistent histological abnormalities in CD patients, this phenomenon has been reported also for patients’ adherent to GFD5, 7 and the mechanism involved is unclear. A HLA-DQ2 gene dose effect has been reported to affect the risk of development, the heterogeneity of clinical presentation and the overall severity of CD2, 13–15 and this effect may also influence histological response to GFD.14 Concomitant conditions may also be important to explain persistence of histological abnormalities not related to gluten toxicity16 and H. pylori infection may be particularly important17 because of the high prevalence in the general population.

Beside adherence and improvement of symptoms during GFD,2 CD related serology is also of little help in predicting mucosal healing during GFD, and as much as 60% of patients with negative tissue-transglutaminase antibodies (t-TG) may have villous atrophy at duodenal histology.18–20 These observations suggest that duodenal histology may be the necessary tool for assessing efficacy of GFD and support the need of further information on the extent and on the factors influencing mucosal healing during GFD.

The aim of our study was twofold: first to assess the effect of GFD on duodenal histology and on clinical and serological correlates; secondly to assess factors affecting histological response to GFD and the effect of HLA gene-dose and of concomitant HP infection on the phenomenon of persistent of intraepithelial lymphocytosis during GFD. To achieve these aims with sufficient statistical power, we carried out a retrospective cohort study involving a large number of adult CD patients followed up with control duodenal biopsies obtained after 12 to 222 months GFD.

Materials and methods

Patient recruitment

A data base of all CD patients seen at our institution since 1990 was compiled to include predetermined information on clinical, laboratory and histological characteristics. From the list of 950, we identified patients older than 14 years at diagnosis, with positive CD-related serology and compatible duodenal histology at diagnosis and with control duodenal biopsies obtained during GFD. The clinical presentation at diagnosis was classified by the presence of intestinal symptoms, extra-intestinal symptoms or as prompted by screening in CD-associated diseases or by family screening.

An educational programme on GFD for newly diagnosed CD patients was provided by our Institutional Dieticians on three sequential visits within the first 6–12 months after diagnosis. Adherence to GFD was assessed at the time of control duodenal biopsy and graded according to a 4-point Likert scale including no dietary digression, 1 serving with gluten per month, 2–4 servings per month or no diet at all.

Serology and histology

CD-related serology consisted of serum IgA-antigliadin antibodies (AGA), IgA-anti-tissue-transglutaminase antibodies (t-TG) and antiendomysial antibodies (EMA) depending on availability of these tests at the time of diagnosis. AGA and t-TG were assayed on enzyme linked immunosorbent assay commercially available kits from Eurospital (Trieste, Italy: cut-off value >7 U/mL for t-TG) or from Pharmacia & Upjohn (Sweden, cut-off value >8 U/mL for t-TG) and EMA were detected by immunofluorescence on primate oesophagus sections from Eurospital. All patients with complete normalization of duodenal mucosa were recalled for HLA typing and all consecutive patients observed during the year 2008 with persistent intraepithelial lymphocytosis during the GFD were also typed for HLA DQA1 and HLA DQB1 using commercial kits (Innogenetics, Ghent, Belgium and Dynal Biotech Ltd., Bromborough, UK).

A minimum four biopsies were collected during upper endoscopy in the second duodenal portion and oriented mucosal-up using filter paper as previously described.21 Intraepithelial lymphocytes have been identified using CD3 immunostaining and a value ≤25 lymphocytes/100 epithelial cells was considered normal.22 Histological changes suggestive of CD have been classified according to the Marsh criteria23 and more recently according to the Marsh-Oberhuber criteria,24 basing classification on the most severe lesion present. Four gastric biopsies have also been routinely obtained, two from the antrum and two from the body of the stomach, in all patients receiving primary care in our Institution.

Expression of results and statistical methods

Histological outcome during GFD has been defined as ‘normalization’ for specimens classified as Marsh 0, as ‘remission’ for specimens with reconstitution of normal villous architecture and with persistent intraepithelial lymphocytosis,5 as ‘no-change’ or as ‘deterioration’ for specimens classified in the same or in a higher Marsh class than at diagnosis. For statistical analysis of sufficient power for assessment of factors affecting histological outcome, patients with ‘normalization’ and with ‘remission’ have been grouped as ‘responsive’ to GFD, and patients with ‘no-change’ and ‘deterioration’ as ‘nonresponsive’.

DQ genotypes were classified in five genotype groups of decreasing risk for CD as described by Margaritte-Jeannin et al.25

Ethical approval was obtained by our institutional ethics committee. Statistical analysis was performed using spss version 10.0 (SPSS Inc, Chicago, IL, USA). All comparisons were made using the Wilcoxon Two-Sample Test, Chi-Square Test and Mantel-Haenszel Chi-Square Test as appropriate. Multiple logistic regression was used according to a backward procedure in order to find the independent variables associated with histological outcome; goodness of fit has been assessed by Hosmer and Lemeshow test.


Baseline characteristics

Four hundred sixty five adult patients (3.3 female/male ratio) were identified, 75% with primary care diagnosis at our Institution. Patients’ characteristic at diagnosis are detailed in Table 1. Baseline clinical, serological and histological characteristics were similar for female and male patients except for higher BMI in males (median 22.6, range 13.6–29.3) than in females (median 20.1, range 15.1 – 35.0, P < 0.0001). Marsh III histological lesion was present in 429 (92%), Marsh II in 25 (6%) and Marsh I in 11 (2%) cases at diagnosis.

Table 1.   Clinical and histological characteristics of coeliac patients at diagnosis and during GFD (BMI = body mass index; IBS = irritable bowel syndrome; IDDM = insulin dependent diabetes mellitus; *P < 0.0001 by Wilcoxon Two-Sample Test)
 At diagnosisDuring GFD
n465 465 
Median age (range): years31 (18–81)   
Median BMI (range)20.5 (13.6 34.2) 21.7 (13.1–35.3)* 
Clinical presentation
 Intestinal symptoms208455257
 Chronic diarrhoea51   
Extra intestinal symptoms130304243
Screening in associated diseases6414  
 Dermatitis herpitiformis42   
Family screening6314  
Associated diseases
  Autoimmune thyroiditis56   
  Dermatitis herpitiformis42   
Duodenal biopsies
 Marsh n465 465 
Marsh-Oberhuber321 321 
Median duration of GFD (range): months  16 (13–222) 
Adherence to GFD
 Yes  39986
 No  6614
CD-related serology

Effect of gluten-free diet

Results are summarized in Table 1. Three hundred and ninety three patients (85%) adhered to the educational programme on GFD run by our Institutional Dieticians. Mean duration of GFD prior to follow-up duodenal biopsies was 16 months (range 13–222 months). Strict adherence to GFD at the time of control duodenal biopsy was reported by 86 % of patients. Complete remission of symptoms during GFD was reported by 75% of the 208 patients originally complaining of gastrointestinal symptoms, by 68% of the 130 complaining of extra-gastrointestinal symptoms, by 54% of the 64 identified by screening of ‘at-risk’ groups and by 64% of the 14 symptomatic patients among the 63 identified by family screening.

CD-related serology was available as t-TG in 365 patients, as EMA in 66 patients and as AGA in 34. CD-related serology was negative in 405 patients (87%) with no difference between male and females.

Histological outcome

In all, 92 % of control duodenal biopsies were obtained in our Institution and assessed by a single pathologist (VV). Three hundred and thirty eight patients (73%) were ‘responsive’ to GFD, including 38 (8%) with ‘normalization’ of duodenal histology and 300 (65%) with ‘remission’ (Figure 1). The number of patients with ‘normalization’ increased marginally from 38 (8%), as based on the cut-off value for intraepithelial lymphocytosis of 2522 adopted for our study, to 41 (9%) and to 46 (10%) by adopting higher cut-off values of 3023 and 4024 respectively, frequently used in clinical practice.

Figure 1.

 Histological response to gluten-free diet in coeliac patients. Remission is defined as reconstitution of villous architecture with persistence of intraepithelial lymphocytosis.

One hundred and twenty seven patients were ‘nonresponsive’ to GFD (27%) including 121 (26%) with ‘no-change’ and 6 (1%) with ‘deterioration’ of duodenal histology. The proportion of ‘responsive’ patients was irrespective of clinical presentation at diagnosis.

The histological outcome during GFD as a function of Marsh classification at diagnosis was as follows: 35 (8%) of the 429 patients classified as Marsh III at diagnosis had histological ‘normalization’ and 300 (70%) had ‘remission’, i.e. reconstitution of villous architecture and persistence of intraepithelial lymphocytosis (Figure 2); ‘no-change’ was observed in 94 patients (22%), but severity of atrophy decreased from class IIIC to IIIA in 28 (41%) in the subgroup of 69 for whom Oberhuber classification was also available. Only one (4%) of the 25 patients in Marsh class II at diagnosis had histological ‘normalization’, and ‘no-change’ was observed in 24 (96%). Two (18%) of the 11 patients in Marsh I class at diagnosis had histological ‘normalization’, ‘no-change’ was observed in 3 (27%) and ‘deterioration’ was observed in 6 (55%), 5 to Marsh II and 1 to Marsh III class.

Figure 2.

 Persistent intraepithelial lymphocytosis with normal villous architecture during gluten-free diet. Panel A: haematoxylin-eosin stained duodenal biopsy; panel B positively stained intraepithelial lymphocytes using CD3 antibodies.

‘Responsive’ vs. ‘nonresponsive’ patients

The 38 patients classified as Marsh 0 were significantly younger than the 427 with persistent abnormalities (median age 26 vs. 31 years, P = 0.0065), but gender distribution (P = 0.3278), clinical presentation (P = 0.4469), presence of associated disease (P = 0.3782), Marsh class distribution at diagnosis (P = 0.3590), length of GFD (P = 0.1696), adherence to GFD (P = 0.1423) and symptomatic response to GFD (P = 0.4656) were similar in the two groups.

At univariate analysis (Table 2) patients ‘responsive’ to GFD were predominantly male, younger and with Marsh III lesions at diagnosis, adherent to GFD and with negative CD-related serology during GFD compared to patients ‘nonresponsive’ to GFD. In the subgroup of 365 for whom t-TG was available at the time of control duodenal biopsies, t-TG tested negative in 44 of 56 patients with persistent Marsh III lesion, but severity of villous atrophy decreased from Oberhuber IIIC to IIIA class in 23 of the 44 patients.

Table 2.   Differences between patients ‘responsive’ and ‘nonresponsive’ to gluten-free diet by histological criteria
 Histological outcomeUnivariate analysisMultivariate analysis
ResponsiveNon-responsivePPoint estimates95% confidence limitsP
  1. * Chi-Square.

  2. † Wilcoxon Two-Sample Test.

  3. ‡ Mantel-Haenszel Chi-Square.

Age at diagnosis
 Median (range): years30.6 (14–81) 33.3 (16–68) 0.0257†   
Clinical presentation
 Intestinal symptoms1467062300.1676*  
 Extra intestinal symptoms96743426    
 Screening in associated diseases45701930    
 Family screening51811219    
Associated disease
Marsh at diagnosis
Dietician educational programme
Adherence to GFD
Length of GFD (months)
 Median (range): months16 (13–222) 15 (13–201) 0.7896‡  
Serology on GFD

At multivariate analysis, (Table 2) male gender, Marsh III lesion at diagnosis and adherence to GFD were independently associated with a histological ‘responsive’ pattern to GFD.

HLA genotype and HP infection in ‘responsive’ patients

HLA genotyping was available in 78 patients ‘responsive’ patients, 28 achieving ‘normalization’ and 50 achieving ‘remission’ of histological lesions during GFD (Table 3). The G3 genotypic group was the most prevalent (37%) followed by G1 (33%), G2 (18%), G5 (7%) and G4 (5%). There was no significant difference in the distribution of G1 (29% vs. 36%), G2 (21 vs. 16%), G3 (39% vs. 36%), G4 (4% vs. 6%) and G5 (7% vs. 6%) genotypic groups between patients achieving histological ‘normalization’ and ‘remission’ during GFD, both by analysing G1 alone (P = 0.504) and the combination of G1 + G2 vs. the other genotypic groups (P = 0.865).

Table 3.   Distribution of HLA DQ genotypes and of the genotypic groups between coeliac patients with ‘normalization’ and with ‘remission’ of duodenal lesions during gluten-free diet. Chi-Square Test P = 0.504 for G1 vs. G2-G5; P = 0.865 for G1 + G2 vs. G3–G5 H1 = DQA1*0501 DQB1*0201 H2 = DQA1*0201 DQB1*0202 H3 = DQA1*0505 DQB1*0301 H4 = DQA1*0301 DQB1*0302 H5 = others
Genotypic groupDQ genotypesAllNormalizationHistologicalRemission95% exact confidence interval
n%n%95% exact confidence intervaln%
All 78 28 25–4850 52–75
H1/H17 0  7  
H1/H219 8  11  
H2/H314 6  8  
H1/H39 3  6  
H1/H42 1  1  
H1/H518 7  11  
H2/H21 0  1  
H2/H43 1  2  
H4/H54 1  3  
H5/H51 1  0  

Results of HP infection were available at follow-up gastric biopsies for 31 patients with ‘normalization’ and 280 with ‘remission’ of duodenal histology and the prevalence of HP infection was 55% (n = 17) and 27% (n = 76) in the two groups (P = 0.0031), respectively.


Our study shows that complete recovery of duodenal mucosa during GFD in CD patients diagnosed at an adult age occurs in a minority of cases (8%) with the vast majority of patients achieving ‘remission’ (65%) and a substantial proportion showing ‘no-change’ (26%) or ‘deterioration’ (1%) of duodenal histology (Figure 1). These findings are consistent with other reports indicating incomplete recovery of intestinal mucosa during GFD,3–10, 14 but the larger size of our cohort, that represents an expansion of a smaller cohort of 72 patients contributed to a previous multi-centre study,9 allows for the necessary statistical power to assess for factors affecting outcome of GFD between patients ‘responsive’ and ‘nonresponsive’ to GFD (Table 2). We are well aware that duodenal specimens are unlikely to be representative of events in the whole small intestine of CD patients, but recent results obtained with capsule endoscopy indicating that most residual changes during GFD remain in the duodenum confirm26 the validity of studies based on duodenal biopsies.

In our study, male gender was independently associated with ‘responsive’ outcome of GFD. This is a new finding and it is tempting to speculate that genotype-gender interactions may be involved as reported for differences in prevalence and clinical phenotypes of several diseases, particularly in the case of autoimmune diseases27 including CD.28, 29 Thus, for example, Cerutti et al.30 have reported a nearly twofold greater risk of CD in diabetic girls than in boys in a large Italian multi-centre study. This phenomenon of genetic dimorphism may theoretically contribute not only to differences of clinical phenotype but also to gender-related differences in treatment outcome.

Age at diagnosis was not independently associated with response to GFD within our cohort of adult patients. Other studies5, 9, 31 have reported, although not consistently,10 higher proportion of complete mucosal healing in children than in adults, a finding consistent with an age-dependent effect on mucosal recovery likely to reflect differences in exposure time to gluten. Taken together, these observations suggest that the detrimental effect of exposure time to gluten on mucosal healing increases progressively and plateaus off at adult age, mirroring a similar plateau effect on the risk of development of autoimmune diseases.32–34 Both these effects suggest that earlier is better for diagnosing CD and support a liberal use of CD-related serology for screening for CD in a variety of clinical conditions.35

Not surprisingly, Marsh class at diagnosis was independently associated with ‘responsive’ outcome, but this was because the best part of our patients ‘responsive’ to GFD were those in ‘remission’, i.e. those in Marsh III class at diagnosis.

Adherence to GFD was, in our study, the factor with the strongest independent association with ‘responsive’ histological outcome, but as in other studies,36, 37 a large proportion (100/127) of ‘nonresponsive’ patients reported strict adherence to GFD. Interestingly, 39/300 patients in ‘remission’ reported poor adherence to GFD, and this may justify their lack of progression towards complete mucosal healing with persistence of intraepithelial lymphocytosis.

As a result of many patients achieving ‘remission’ and of many other with ‘no-change’ in Marsh I and II class, a large proportion of patients (71%, Table 1) had persistent intraepithelial lymphocytosis during GFD and the reason behind lack of progression towards complete mucosal healing remains unclear. Food contamination or inadvertent gluten ingestion may play a role by causing the threshold daily dose for gluten toxicity to be exceeded. This threshold dose may lie somewhere between 10 and 50 mg/day gluten8, 38, 39 and Italian patients may introduce 6–8 mg gluten daily given a consumption of nearly 400 g of ‘gluten-free’ products40 containing 20 mg gluten/Kg. This value of 6–8 mg gluten per day is close to the threshold 10 mg dose and this safe dose can easily be exceeded by inadvertent gluten consumption thus causing persistent intraepithelial lymphocytosis, a sensitive marker of gluten consumption.8, 41

A novel aspect of our study is that we assessed whether genetic or environmental factors might be involved to explain persistent intraepithelial lymphocytosis in patients adhering to a GFD. In relation to genetic factors, HLA-DQ2 homozygous antigen presenting cells have been reported to stimulate in vitro T-cell clones responses to gluten peptides42 more efficiently than heterozygous cells, a gene-dose effect potentially making homozygous patients more vulnerable to ingestion of trace amounts of gluten and at risk of complications.2 This gene-dose effect was not confirmed by our study. Despite a trend for the G1 genotypic group to be more represented in patients with ‘remission’ (69%) than in those with ‘normalization’ (31%, Table 3), this difference was not statistically significant even by combining results of G1 with G2 genotypic groups. i.e. the two genotypic groups carrying the highest risk for CD in the Italian population.25

Beside genetic factors, concomitant environmental factors16 may cause intraepithelial lymphocytosis unrelated to CD, including HP infection.17 Our study has clearly ruled out HP infection as an environmental confounding factor17 for persisting intraepithelial lymphocytosis in CD patients given the higher prevalence of HP infection in patients with ‘normalization’ (55%) than in those in ‘remission’ (27%).

A further possibility to explain persistent intraepithelial lymphocytosis during GFD is to hypothesize a role for persistent activation of the innate immune system that is thought to be important in the pathogenesis of CD.15, 43 Amino-terminous peptides of alpha gliadin that do not activate adaptative immunity via HLA-DQ restricted mucosal T cells have been reported to stimulate cytokine IL-15 over-expression by epithelial cells and to cause lymphocytic intraepithelial infiltration and epithelial damage.15 Whether trace amount of gluten present in ‘gluten-free’ products may be sufficient to activate the innate immune system is speculative, but theoretically plausible by assuming a lower activation threshold for innate than for adaptive immunity.

An aspect of clinical interest of our results is that not only symptomatic improvement and adherence to GFD but also normalization of CD-related serology is not a guarantee of improvement of duodenal mucosa as reported by others2, 8, 11, 18, 19 thus underlining the need for a control biopsy after a minimum 1 year GFD2 for identification of patients ‘nonresponsive’ to GFD and at risk of developing severe complications.11

In summary, our study indicates that reversal to normality of small intestinal histology is rarely achieved in CD patients during prolonged and strict adherence to GFD and despite remission of symptoms. In a majority of patients, normal villous architecture is restored, but intraepithelial lymphocytosis persists independently of a HLA gene dose-effect and of HP infection. Studies on patients with type-1 refractory CD44 and on patients developing severe complications despite clinical improvement11 suggest a potential of persisting mucosal abnormalities for development of severe CD-related complications. Given the observation that negative CD related serology and symptoms disappearance are not reliable predictors of histological ‘responsive’ outcome, control duodenal biopsies are to be recommended11 in adult patients to identify those at risk of complications.


Declaration of personal and funding interests: None.