Potential Celiac Children: 9-Year Follow-up on a Gluten-containing Diet

Renata Auricchio MD, PhD; Antonella Tosco MD; Emanuela Piccolo MD; Martina Galatola PhD; Valentina Izzo PhD; Mariantonia Maglio PhD; Francesco Paparo PhD; Riccardo Troncone MD, PhD; Luigi Greco MD, PhD


Am J Gastroenterol. 2014;109(6):913-921. 

In This Article



From 2001 to 2010, 210 children (141 girls; median age: 6.4 years; range: 1.1–16.9 years) were potential celiac patients because of two or more positive anti-TG2 antibody tests, confirmed by EMA positivity, and normal small duodenal mucosa (91 with Marsh grade 0 and 119 with Marsh grade 1 sec. Oberhuber; Figure 2).

Figure 2.

Photograph of a typical duodenal mucosa of a potential celiac disease (CD) child with normal architectural mucosa and increased intraepithelial infiltration.

Not all patients agreed to undergo a new biopsy every 2 years, but we could not exclude them, to avoid a selection bias. Nevertheless, a second biopsy was performed in 109 cases (72 children underwent 1 biopsy, 17 underwent 2 biopsies, and one child underwent 3 biopsies during the follow-up).

During follow-up, gluten intake was also monitored by a nutritionist with a food frequency questionnaire. The amount of gluten intake (15 g per day) was comparable to that of a control population matched for age and sex.

Ninety-three of 210 (44.2%) patients belonged to the "at-risk" group: 59/93 were first-degree relatives of celiac patients, and 34 were affected by other autoimmune diseases (20 type 1 diabetes, 13 thyroiditis, and one with both type 1 diabetes and thyroiditis).

In the same period of time, we diagnosed 1,800 celiac cases on the basis of the European Society of Gastroenterology, Hepatology and Nutrition 1999 criteria. The global rate of potential/confirmed CD was 11.6%.

At entry, 35 patients started a gluten-free diet because of symptoms or parents' choice and were excluded from the study; 3 of them resumed a gluten-containing diet because of no improvement of the symptoms, but they were not re-admitted to the study cohort. One hundred and seventy-five asymptomatic children were left on a gluten-containing diet (Figure 1) and entered the prospective longitudinal follow-up study (median: 54 months, range: 6–108). One hundred and nine children underwent more than one duodenal biopsy during the observation period.

During follow-up, 22 patients were put on a gluten-free diet because of subjective or objective symptoms at a median of 12 months (range: 6–60 months) after enrollment (Figure 1). One hundred and fifty-three asymptomatic children continued on a gluten-containing diet, with a minimum gluten intake of 15 gr per day, and were monitored at each checkup. All patients had normal nutritional parameters (iron assessment, albumin). Autoimmune thyroiditis developed in five patients. Over 9 years, 23 children developed Marsh stage 3 CD in the duodenal mucosa (Figure 1).

Individual Profile of Anti-TG2 Antibodies

During follow-up, 66 (43%) patients showed persistently elevated anti-TG2, 30 (20%) became negative after at least three positive values, and 56 (37%) showed a fluctuating anti-TG2 course with transiently negative values.

The pattern of EMA positivity was parallel to the anti-TG2 value, but because of the coarseness of the EMA titration scale anti-TG2 values were preferred, in order to monitor patients by an accurate, objective scale. EMA level was correlated to anti-TG2 level (Pearson's r=0.223, P<0.001). Anti-TG2 antibodies did not show any significant correlation with the age at enrollment (Pearson's r=−0.07, P=0.22), age at any later time, or sex. The anti-TG2 titer at entry was not statistically different between those who remained potential CD patients and those who progressed (become atrophic or developed symptoms) (M±s.d.: 22.86±29.57 vs 29.97±37.85 P=0.098), but the variation of anti-TG2 was correlated to the final outcome. Table 1 shows the flow of anti-TG2 by outcome at the final biopsy. None of the 'negative' anti-TG2 patients developed full-blown celiac, whereas among those who developed severe damage 77.8% had persistently positive anti-TG2, compared with 43% of those who did not develop damage (χ 2=8.56; P=0.014). The median value of anti-TG2 at first diagnosis (anti-TG2 T0) was higher in the persistent (31.4) cases than those who became negative (16.9) or fluctuated (20.0; analysis of variance F=3.29; P=0.04). A large spread was observed in the group of patients with persistent positivity for anti-TG2 (data not shown). The value of anti-TG2 at time 0 was not significantly different across the HLA risk groups (F=0.8, P=0.5).

CD Cumulative Incidence

The survival functions up to 9 years, where "event" was defined as the development of severe mucosal lesions, shows that at 3 years 144 (86%) cases remained potential CD, at 6 years 138 (73%) remained potential, and at 9 years 133 (67%) still had a normal duodenal architecture (Figure 3a and Supplementary Table S1 online).

Figure 3.

Cumulative survival function in the cohort study. (a) Percentage of cases who remained potential during follow-up. (b) Cumulative survival function by sex. (c) Progression to mucosal damage by the presence or absence of anti-tissue-transglutaminase (anti-TG2) deposits at entry.

Boys more frequently developed severe mucosal damage compared with girls (Figure 3, panel b). At 9 years, 46% of boys remained potential, compared with 78% of girls (Gehan Wilcoxon=4.097, P=0.043).

Survival was stratified by HLA risk groups according to Bourgey et al.:[19] a clear rising trend of event-free survival from higher- to lower-risk classes (double DQ2=77%, DQ2 in trans=80%, single DQ2=84%, DQ8=88%, no DQ2/no DQ8=100%) was observed. However, this trend was not statistically significant owing to the relatively small size of the groups (log-rank χ 2=0.53).

Intestinal anti-TG2 antibody deposits in the mucosa were more frequent in cases that progressed to mucosal damage (82.6%) than in those remaining potential CD (64.1%), with marginal statistical significance (χ 2=3.08, P=0.06; Table 2 and Figure 3, panel c; Wilcoxon=3.83, P=0.05).


We compared 11 candidate-gene SNPs (Supplementary Table S2 online) between the patients who remained potential and those who developed severe mucosal damage, as previously described.[20,21] Despite the lack of power owing to the small sample size, 2 of 11 SNPs showed a significant difference between persisting potential CD and those who developed atrophy (Table 3). The SNP in the gene RGS1, whose main function is the homing of intraepithelial lymphocytes, had an AA genotype less frequently in those who developed CD compared with those who remained potential. On the other hand, the IL12A SNP genotype AA was twice as common in celiac patients compared with potential patients. Other differences in candidate-gene SNPs were also observed but did not reach statistical significance, namely the c-REL SNP genotype AA was 16.6% less frequent in those who developed CD.

Discriminant Analysis

Among the 175 cases of the study cohort, we attempted to identify those variables that helped discriminate at entry (time 0) the 23 cases who developed CD from the 152 who remained potential. In the cohort of 152 cases, the 22 cases who started a gluten-free diet during follow-up were included, as they had the proper entry criteria and none of them developed small intestinal atrophy.

γδ T-lymphocytes and CD25 infiltration at the starting biopsy appeared to be good predictors of outcome ( Table 4). Sex, as we observed previously, also contributed to the discrimination, with boys being more at risk. The genotypes of three of the analyzed SNPs (in IL12A, IL2/IL21, and OLIG3) also contributed significantly. It is interesting to note that, although the IL12A SNP already showed significant differences in genotype frequency the IL2/IL21 SNP was selected by the procedure after having considered the other candidate genes.

Finally, the value of anti-TG2 at entry also contributed to the discriminant equation. Intestinal deposits of anti-TG2 at entry (time 0) did not contribute significantly to the discriminant equation after we considered more powerful risk factors.

The contribution of each variable to lower the Wilks' lambda (discriminating capacity) was small but significant for each variable. Altogether, they combine to create a profile of the cases with unfavorable outcome, compared with those who persisted with normal mucosa. By multiplying the canonical discriminant coefficients by each of the selected variables and adding a constant term, we obtained the discriminant score for each individual, which provided the probability of membership in one of the two groups. By assigning to each individual the highest probability of membership among the potential CD patients who kept a normal mucosa vs those who developed severe damage, we estimated the efficiency of the discriminating equation. Overall, 74% of cases were correctly predicted. Indeed, 75% of cases who did not develop damage were correctly predicted at time 0 by this equation, whereas 34% of those who developed full CD were wrongly classified as still potential. Keeping in mind the uncertainty of predicting the outcome in this domain, this classification may appear overoptimistic, as it is based on the equation obtained by the same cases that had classified at that time, but when a cross-validation was applied, by classifying each case through an equation that did exclude the case under evaluation (jackknifing procedure), the results were indeed quite stable: 73% of still-potential patients and 65% of those who became celiac patients were correctly classified, supporting the robustness of the model (Table 5). We could not validate this model any further owing to the small size of the cohort with children who developed mucosal damage.