What Is Known
- The cumulative percentage of children with celiac disease who normalize antitissue transglutaminase after 18 months is 64%.
- No literature exists examining multiple independent predictors of antitissue transglutaminase over time in children with celiac disease.
What Is New
- Lower baseline antitissue transglutaminase at diagnosis is an independent predictor of earlier normalization.
- Patients with type 1 diabetes mellitus are less likely to normalize antitissue transglutaminase levels, and have longer time to normalization.
- Further research and education for these higher-risk populations is needed, including an examination of the utility of multidisciplinary celiac care.
Celiac disease (CD) is an autoimmune reaction to gluten, leading to intestinal inflammation, villous atrophy, and malabsorption (1,2). The prevalence of CD is 1% in the general population (2,3), making it the most common autoimmune gastrointestinal disorder. CD is triggered by the ingestion of gluten present in cereals such as wheat, barley, and rye in genetically susceptible individuals (1,4). In 1997, tissue transglutaminase (tTG) was discovered as the autoantigen in CD, leading to the development of tests to identify anti-tTG (atTG) antibodies as a screen for CD. The most commonly used method is enzyme-linked immunosorbent assay (ELISA) based on atTG-immunoglobulin A (IgA), which has a good sensitivity (90%–98%) and specificity (95%–97%) for CD (5–9). High levels of atTG are more likely to be associated with villous atrophy (5), with Marsh score representing the histologic severity of intestinal involvement in CD (10). A life-long gluten-free diet (GFD) is an effective treatment to alleviate symptoms, normalize atTG, and heal the intestinal mucosa in patients with CD (4,11,12).
In North America patients with CD are typically diagnosed based on positive screening atTG testing, followed by endoscopy with characteristic histopathology for confirmation (3,12). Monitoring for treatment response in all patients with CD includes following the atTG titers over time, with a goal of a normal atTG level on a GFD. The North American Society of Pediatric Gastroenterology, Hepatology and Nutrition (NASPGHAN) guidelines use expert opinion to recommend following atTG titers every 6 months after initiating a GFD until normalization, and then on a yearly basis (3). Normalization of atTG on repeat testing suggests compliance with the GFD (3), and may obviate the need for repeat mucosal biopsy in the majority of patients (13). A delay in resolution of atTG titers despite reported compliance with a GFD may also be an indication for intestinal biopsy (4), leading to increased exposure to endoscopic procedures.
The time to normalization of atTG titers after adoption of a GFD is clinically relevant, and is a common question and concern of patients with CD and their families. Variability, however, exists in the current literature regarding time to atTG normalization, and pediatric data are limited. Studies suggest resolution of symptoms and intestinal damage within weeks to a few months of adherence to a GFD (5,14), with time to normalization depending upon the pretreatment titers (9). Hogen Esch et al (15) showed that 64% of pediatric patients had normalized atTG by 18 months on a GFD (15), whereas other pediatric studies reported a trend toward normalization by 6 months but lacked more specific normalization times (4,9,16). With respect to independent predictors of time to atTG normalization, Rajani et al (17) noted that more Caucasian children with CD had a normal atTG level compared with South Asians (SAs) after 1 year on a presumed GFD (64% vs 29%; P < 0.001); however, additional investigation into other clinical predictors that may affect atTG normalization in children is still required (16).
Based on a review of the literature to date, we were unable to identify a study exclusively examining the trend and predictors of atTG over time in patients with CD in a pediatric population. Therefore our primary goal was to evaluate time to normalization of atTG in our local pediatric CD population postdiagnosis, and to examine for independent predictors that may affect the duration to atTG normalization. The results could then be used to improve our understanding of the natural history of atTG titers after diagnosis and initiation of a GFD in children with CD, and guide optimal timing for consideration of repeat endoscopy in reported GFD compliant patients who exhibit prolonged atTG elevation. In addition, identifying independent predictors of prolonged atTG elevation may inform clinical practice by aiding clinicians in providing guidance to higher-risk patients and families at the time of CD diagnosis, and during follow-up.
Institutional review board ethics approval was obtained before beginning the study (HREB: Pro00053480). All known patients with a diagnosis of CD in the Stollery Children's Hospital Celiac Clinic Patient Database (Edmonton, Alberta, Canada) from January 2007 up to May 2014 were reviewed for inclusion. Exclusion criteria from the main analysis were atTG <7 units per milliliter (U/mL), IgA deficiency, patients with <1 follow-up atTG result, and patients with a serologic diagnosis of CD (defined as patients with a baseline atTG >200 U/mL that did not undergo endoscopic biopsy confirmation). Serologic diagnosis patients were, however, included in a second Cox regression analysis described below.
A retrospective chart review was completed to evaluate atTG normalization time in pediatric patients with CD (younger than 18 years). Data were collected on the following clinical predictors to assess for effect on time to atTG normalization: age at diagnosis, sex, ethnicity, initial atTG, Marsh score at diagnosis, GFD compliance (GFDC), medical comorbidities, and a family history of CD. The available atTG measurements over time were recorded for each patient to assess for time of normalization.
We defined normalization of atTG as <7.5 U/mL based on local laboratory standards (≥7 U/mL is a positive laboratory result, with the coefficient of variation for the test in our laboratory being 4%–5%) (18). The chart notes from in-person interviews between an expert dietician and the patients and caregivers were used to assess compliance with the GFD. Adequate time of follow-up was defined as ≥2 years from diagnosis. For patients who did not normalize atTG, but who had follow-up or data limited to <2 years from diagnosis, they were considered loss to follow-up rather than failure to normalize. The date of diagnosis was defined as the date of endoscopy in the biopsy-confirmed patients with CD, and this was also used as the assumed date of initiation of the GFD. For the serologic diagnosis patient subgroup, the date of baseline atTG was used as the date of diagnosis.
Means and standard deviations were used for analysis of descriptive data. Kaplan-Meier survival analysis was conducted to calculate median time to atTG normalization. Assessment of significance for Kaplan-Meier survival analyses was obtained by the log-rank test to determine P values. An overall log-rank test was conducted for the ethnicity Kaplan-Meier survival analysis, followed by pairwise log-rank comparisons to determine which ethnic groups had differences in survival distributions. Univariate regressions were calculated to examine the significance between predictor variables and time to atTG normalization. Variables with significance level of P < 0.20 and any clinically significant variables were assessed for inclusion in the Cox regression. The final Cox regression model included the independent variables of age at diagnosis, ethnicity, sex, type 1 diabetes mellitus (T1DM), family history, Marsh score at diagnosis, initial atTG level, and GFD compliance. Variables with significance level of P < 0.05 through were considered statistically significant. The initial atTG level was scaled down for Cox regression by a factor of 1000. The frequencies of other medical comorbidities found in the patient population examined were too small in number to be assessed in Cox regression (hypothyroidism, Down syndrome [DS], autoimmune hepatitis, eosinophilic esophagitis, and granulomatosis with polyangiitis). Assumptions of hazard proportionality were checked, and no violations were found. Patients without adequate follow-up were included in the data analysis as censored cases.
A second Cox regression analysis including the serologic diagnosis patients was completed to assess for effect of baseline atTG and Marsh score on time to normalization, given that these patients have an expected higher baseline atTG level at diagnosis. Given that serologic diagnosis patients did not undergo biopsy confirmation and therefore had no associated Marsh score, we assigned a higher hypothetical Marsh score of 4 at diagnosis for each of these patients as the literature suggests that the histologic severity of intestinal involvement correlates to the level of the titer for atTG testing.
Of the 616 patients reviewed 487 patients met inclusion criteria (Fig. 1). An additional 45 serologic diagnosis patients were included in a second Cox regression analysis. Mean age was 9.3 years (standard deviation 4.1) at diagnosis (range 1–17 years), with 64% girls. The minimum patient follow-up time was 6 months, and the maximum time of patient follow-up was 6 years. The distribution of patient ethnicity was 61% Caucasian, 14.4% SA, 2.1% Middle Eastern, 1.4% East Asian, 0.4% First Nations, and 20.7% unspecified. For patient comorbidities, 44 (9%) had T1DM, 13 (2.7%) had autoimmune hypothyroidism, 8 (1.6%) had DS, and 5 (1%) had eosinophilic esophagitis. The average baseline atTG for biopsy confirmed patients was 446.3 U/mL, and 841.9 U/mL for serologic diagnosis patients. Compared to patients without T1DM, children with T1DM had significantly higher median baseline atTG levels (725.4 vs 420.1 U/mL, P = 0.005) and lower GFD compliance (77.3% vs 89.1%, P = 0.021).
Time to Antitissue Transglutaminase Normalization
Approximately 80.5% of patients (n = 392) normalized atTG levels within the study time period. Kaplan-Meier survival analysis of time to atTG normalization (Fig. 2) showed median time to normalization was 407 days for all patients (95% confidence interval [CI] [361–453]). Twenty-five percent of all patients had normalized by 240 days (standard error 11 days), and 75% of patients by 907 days (SE 93 days). For GFD compliant patients, median time to normalization was 364 days (95% CI [335–393]). Twenty-five percent of GFD compliant patients had normalized at 223 days (SE 9.6 days), and 75% at 590 days (SE 39.9 days). Patients with T1DM took significantly longer to normalize (Fig. 3), with a median time of 1204 days (95% CI [199–2209], P < 0.001, log-rank test). Ethnicity was associated with longer time to normalization (Fig. 4; P = 0.001, overall log-rank test), with SA patients taking a median 809 days (95% CI [262–1356], P < 0.001, pairwise log-rank test); however, the validity was limited due to wide difference in censoring between ethnicities (Caucasian 18.2% vs SA 35.2%). Sex, family history of CD, and higher Marsh score at diagnosis did not show statistically significant differences in time to normalization with different percentages of censored cases between groups.
Independent Predictors of Antitissue Transglutaminase Normalization
Cox regression demonstrated T1DM (hazard ratio [HR] = 0.36 [0.24–0.55], P < 0.001) and higher baseline atTG (HR = 0.52 [0.43–0.63], P < 0.001) were significant predictors of longer time to atTG normalization. Conversely, GFDC was a significant predictor of earlier normalization (HR = 13.91 [7.86–24.62], P < 0.001). Age at diagnosis, sex, family history of CD, ethnicity, and Marsh score at diagnosis were not significant independent predictors of time to atTG normalization. An additional Cox regression analysis was completed including serologic diagnosis patients to assess for an effect of baseline atTG and Marsh score on atTG normalization, and no changes in outcomes were noted.
Quality improvement and identification of modifiable risks in clinical care is paramount to ensuring the best outcomes for patients. The results of our study demonstrate that GFD compliant patients with CD take a median time of 1 year to normalize atTG levels postdiagnosis, and that compliance with the GFD and lower atTG at diagnosis are independent predictors of earlier atTG normalization in keeping with expected results. Conversely, T1DM and SA patients had longer atTG normalization times. Importantly, our study population was consistent with expected characteristics for the pediatric CD population, including the majority being girls and a larger proportion of affected individuals being of Caucasian and SA descent (2,3,16). In addition, T1DM, autoimmune hypothyroidism, and DS were the 3 most prevalent concomitant medical conditions noted in our patient population, fitting with the literature of patients more at the risk for having CD.
Exclusion of gluten from the diet should result in a gradual decline in serum atTG levels (9). A wide variation of rate and time to atTG normalization in children with CD was noted. The median normalization time of 407 days for the 80.5% of patients that normalized within the study time frame, and 364 days for GFD compliant patients. In a retrospective study of 129 pediatric patients with CD, Hogen Esch et al (15), however, showed the cumulative percentages of children on the GFD who became negative for atTG at 6 months, 12 months, 18 months, and 2 years was 35%, 55%, 64%, and 78%, respectively. Other smaller pediatric studies reported a trend toward normalization by 6 months in GFD compliant patients, but did not explicitly note time to normalization for all comers (4,9,16). No studies that included pediatric patients with suboptimal GFDC were found in the literature. Overall our results for normalization times in GFD compliant patients were similar to the findings of Hogen Esch et al (15).
Also in keeping with the anticipated findings, GFD compliant patients were >13.9 times more likely to have atTG normalized than noncompliant patients, and a lower atTG at diagnosis was a significant predictor of earlier normalization. Previous studies have shown that high levels of atTG are more likely to be associated with villous atrophy than low or borderline levels (5). Donaldson et al (10) also noted that the Marsh score, and therefore the severity of intestinal involvement in CD, correlates to the level of the titer for atTG testing. Despite this assertion, our data show the Marsh score at diagnosis was not a significant independent predictor of duration to normalization, even though a significantly longer time to atTG normalization with a higher baseline atTG at diagnosis was noted. Even when including the serologic diagnosis patients in the Cox regression, who had higher baseline atTG values than the biopsy confirmed patients and a higher assigned hypothetical Marsh score of 4, Marsh score at diagnosis was not a significant predictor of longer duration to normalization. We acknowledge the limitations given the hypothetical assumption of Marsh score for the serologic diagnosis patients; however, it is still interesting to note. This may suggest that time to intestinal healing may not be directly proportional to time to biochemical resolution of atTG, which is a proposal that would require further research to assess intestinal mucosal healing in comparison to atTG serum levels over time on a GFD.
Patients with T1DM are 2.8 times less likely to normalize atTG levels, and have a significantly longer time to normalization. Of note, patients with T1DM had higher baseline atTG and worse GFD compliance. These results may suggest that CD is more difficult to control in the T1DM population, who are faced with more complex dietary requirements that may make compliance more difficult. The finding of significantly longer time to atTG normalization in patients with T1DM is, however, still valid, because baseline atTG and GFD compliance have been adjusted in the Cox regression model. This is clinically relevant given that poorly controlled CD can lead to increased difficulty with glycemic control, and therefore increase the long-term consequences of poor metabolic control in diabetics such as microvascular complications (19). Furthermore, patients with T1DM may be at increased risk of continued intestinal inflammation compared with the non-T1DM CD population, thereby increasing the potential risks of continued malabsorption, growth restriction, diminished bone health, and intestinal lymphoma (3). It also highlights the role that a CD-specific multidisciplinary clinic can play in reviewing GFDC, given that many of the patients with T1DM are not followed in a CD-specific clinical setting.
Alternatively, this finding in the T1DM population raises some concern about biochemical overlap in the autoantibodies being detected by the atTG screen, and whether it may be picking up on structurally related antibodies that flare during times of systemic illness or poor diabetic control in patients with known increased autoimmunity. Waisbourd-Zinman et al (20) found that elevated serum atTG antibody levels in patients with newly diagnosed T1DM will normalize in 35% of cases within 1 year, even on a gluten-containing diet, and suggested that these patients did not have true CD. Simell et al (21) reported 49% transient atTG antibody positivity in patients genetically susceptible to CD who were tested periodically during the first 4 to 7 years of life, and suggested that a regulatory immune response blocks the incipient CD despite the transient detectable atTG for young children at risk. Liu et al (22) studied patients with a genetic susceptibility to CD and found that atTG antibody blood levels varied over time, even in patients without biopsy evidence of CD. These studies have shown detectable atTG in non-CD patients that normalized over time despite remaining on a gluten-containing diet, suggesting that in T1DM and non-T1DM patients there may still be atTG detectable that is not related to poor GFDC. One then postulates that a similar yet unknown mechanism may be related to the finding described in our study.
There may be an association between SA ethnicity and longer duration to normalization; however, the validity of this finding on Kaplan-Meier survival analysis was limited with wide differences in censoring, and significance was not found on Cox regression. Rajani et al previously noted more Caucasian CD children had a normal atTG level compared with SAs (64% vs 29%; P < 0.001) after 1 year on a presumed GFD. Despite the limitations in the significance of our results, potential contributing factors to this finding may be increased genetic sensitivity to ambient gluten in the SA population (17), language barriers in GFD teaching, and the risks of imported foods with inadequate labeling or standards for gluten-free products. The apparent delay in atTG normalization suggests that careful follow-up and culturally congruent education should be developed for SA children with CD (17).
There were limitations to the present study. Patients may have had different times to uptake of the GFD from date of diagnosis and GFD teaching, which could account for some differences in time to resolution, which were not specifically accounted for. In addition, some patients may have started a GFD after a positive initial atTG but before biopsy confirmation. The majority of patients with biopsy-confirmed CD, however, underwent endoscopy within 6 weeks from the initial atTG; therefore, the potential effect of an earlier than recorded GFD initiation would be limited. Another limitation is that atTG levels are not tested continuously but instead tested in intervals, and therefore patients may have normalized earlier than documented between these atTG serum testing intervals. For some patients the delay between testing intervals may have lead to overestimation of normalization time, but our median normalization times were in keeping with previous studies and the large sample size aids in limiting this effect. Our method of determining GFDC was based on an in-person dietician review with the family and patient report, and therefore is subject to patient and caregiver error in addition to unknown ambient gluten exposures. This method of GFDC assessment is, however, the criterion standard. Currently there are no validated, readily available biochemical methods of objectively and independently measuring exposure to gluten in the diet, but utilizing such tools in future research would minimize this limitation.
In conclusion, compliance with the GFD and lower atTG at diagnosis are independent predictors of earlier atTG normalization. Patients with T1DM are less likely to normalize atTG levels, and have longer time to normalization. There may be an association between SA ethnicity and longer duration to normalization, but the validity of this finding was limited in our study. The knowledge that patients with T1DM, and possibly the SA population, have longer duration to normalization can help to inform clinical decision-making and minimize exposure to sedation and procedures. Closer follow-up and rigorous dietary review for these patients are key modifiable components in providing improved care given that longer duration to normalization for these patients may put them at increased risk for ongoing intestinal inflammation and the associated clinical consequences. These findings also highlight the importance of a multidisciplinary team approach with a major role for the dietician in managing CD, especially for these higher-risk populations.
1. Lebwohl B, Rubio-Tapia A, Assiri A, et al. Diagnosis of celiac disease. Gastrointest Endosc Clin N Am
2. Rubio-Tapia A, Ludvigsson JF, Brantner TL, et al. The prevalence of celiac disease in the United States. Am J Gastroenterol
3. Hill ID, Dirks MH, Liptak GS, et al. Guideline for the diagnosis and treatment of celiac disease in children: recommendations of the North American Society for Pediatric Gastroenterology, Hepatology and Nutrition. J Pediatr Gastroenterol Nutr
4. Haines ML, Anderson RP, Gibson PR. Systematic review: the evidence base for long-term management of coeliac disease. Aliment Pharmacol Ther
5. Fasano A. A Clinical Guide to Gluten-Related Disorders. Philadelphia, PA: Lippincott Williams & Wilkins; 2014.
6. Dieterich W, Laag E, Schöpper H, et al. Autoantibodies to tissue transglutaminase as predictors of celiac disease. Gastroenterology
7. Hopper AD, Hadjivassiliou M, Hurlstone DP, et al. What is the role of serologic testing in celiac disease? A prospective, biopsy-confirmed study with economic analysis. Clin Gastroenterol Hepatol
8. Hussain S, Sabir MUD, Afzal M, et al. Coeliac disease—clinical presentation and diagnosis by anti tissue transglutaminase antibodies titre in children. J Pak Med Assoc
9. Leffler D, Edwards George JB, Dennis M, et al. A prospective comparative study of five measures of gluten-free diet adherence in adults with coeliac disease. Aliment Pharmacol Ther
10. Donaldson MR, Book LS, Leiferman KM, et al. Strongly positive tissue transglutaminase antibodies are associated with Marsh 3 histopathology in adult and pediatric celiac disease. J Clin Gastroenterol
11. Rostom A, Murray JA, Kagnoff MF. American Gastroenterological Association (AGA) Institute technical review on the diagnosis and management of celiac disease. Gastroenterology
12. Rubio-Tapia A, Rahim MW, See J, et al. Mucosal recovery and mortality in adults with celiac disease after treatment with a gluten-free diet. Am J Gastroenterol
13. Bannister EG, Cameron DJ, Ng J, et al. Can celiac serology alone be used as a marker of duodenal mucosal recovery in children with celiac disease on a gluten-free diet? Am J Gastroenterol
14. Macdonald WC, Brandborg LL, Flick AL, et al. Studies of celiac sprue. IV. The response of the whole length of the small bowel to a gluten-free diet. Gastroenterology
15. Hogen Esch CE, Wolters VM, Gerritsen SA, et al. Specific celiac disease antibodies in children on a gluten-free diet. Pediatrics
16. Rajani S, Huynh HQ, Turner J. The changing frequency of celiac disease diagnosed at the Stollery Children's Hospital. Can J Gastroenterol
17. Rajani S, Alzaben A, Shirton L, et al. Exploring anthropometric and laboratory differences in children of varying ethnicities with celiac disease. Can J Gastroenterol Hepatol
18. Alberta Health Services Laboratory Inquiry. University of Alberta Hospital, Edmonton, Alberta, Canada.
19. Sud S, Marcon M, Assor E, et al. Celiac disease and pediatric type 1 diabetes: diagnostic and treatment dilemmas. Int J Pediatr Endocrinol
20. Waisbourd-Zinman O, Rosenbach Y, Shalitin S, et al. Spontaneous normalization of anti-tissue transglutaminase antibody levels is common in children with type 1 diabetes mellitus. Dig Dis Sci
21. Simell S, Hoppu S, Hekkala A, et al. Fate of five celiac disease-associated antibodies during normal diet in genetically at-risk children observed from birth in a natural history study. Am J Gastroenterol
22. Liu E, Bao F, Barriga K, et al. Fluctuating transglutaminase autoantibodies are related to histologic features of celiac disease. Clin Gastroenterol Hepatol