A Longitudinal Study of the Effects of a Gluten-Free Diet on Glycemic Control and Weight Gain in Subjects With Type 1 Diabetes and Celiac Disease

Between 1994 and 1998, the pediatric diabetic clinic at John Radcliffe Hospital, Oxford, consisted of 230 children who were all seen every 3 months. Screening for celiac disease was performed annually from the first year after diagnosis of type 1 diabetes by obtaining a sample of blood from each patient for the determination of serum IgA EMA and total IgA levels. Serum from subjects with low total IgA levels (<0.5 g/l) was assayed for IgG AGA antibodies.

A total of 10 children were EMA positive and 1 child with low total IgA levels was AGA positive, accounting for 4.8% of the clinic population (10). Of these 11 case subjects, 3 were detected on their first annual screening test. Only one patient demonstrated symptoms typical of celiac disease, presenting at the time of screening with failure to thrive and steatorrhoea. Also, four subjects had retrospective histories of mild abdominal discomfort and diarrhea. All subjects were counseled regarding a jejunal biopsy, which was carried out by endoscopy under sedation. Multiple biopsy specimens were obtained from the distal duodenum and proximal jejunum. The specimens were examined by a hospital pathologist and pediatric gastroenterologist. All 11 case subjects showed classical histological features of celiac disease and were commenced on a GFD under supervision of a dietitian.

At the time of diagnosis of celiac disease, each case was matched for age, sex, and duration of diabetes with two type 1 diabetic control subjects from the same clinic who had negative AGA/EMA status (Table 1).

Height, weight, and HbA_1c were measured on all subjects every 3 months from diagnosis of diabetes, and C-peptide levels were measured every year. These data, together with daily insulin requirements and insulin regimen, were recorded in subject case notes at each clinic appointment. Data were collected retrospectively from the case records of both celiac-positive case subjects and matched control subjects from the time of diagnosis of diabetes to diagnosis of celiac disease, and were collected prospectively from start of a GFD in case and control subjects for 4 years. Once a diagnosis of celiac disease was confirmed and a GFD was started in case subjects, AGA/EMA levels were tested every 3 months until negative and every year thereafter as a marker of good dietary compliance. In control subjects, AGA/EMA levels were tested yearly and to date remain negative. Height and weight were measured by a single observer.

EMAs were detected using indirect immunofluorescence on cryostat sections of lower third primate esophagus. Sera were screened at a 1:5 dilution and were reincubated with rabbit anti-serum IgA-specific fluorescein conjugate (Dako). Sections were examined under UV light immunofluorescence. Positive sections were identified by characteristic perivasicular endomysial staining. Total serum IgA levels were determined by immunochemical quantification (Beckman). IgG AGA were analyzed using rodent stomach tissue cryostat sections preincubated in gliadin. Sera were screened at a 1:10 dilution and incubated on sections at room temperature for 30 min before being washed and reincubated with fluorescinated anti-human IgG conjugate (Dako). Sections were read under UV light immunofluorescence for characteristic staining patterns. HbA_1c was measured by high-pressure liquid chromotography (Diamat). The intra-assay coefficients of variation (CVs) were 1.9 and 2.2% at HbA_1c levels of 6.9 and 11.5%, respectively. The inter-assay CVs were 2.7 and 2.3% at HbA_1c levels of 7.0 and 11.6%, respectively. Normal population range was 4.0–6.5%. C-peptide was measured using a double-antibody radioimmunoassay from DPC Ltd. The intra-assay CV at C-peptide levels of 294 and 2,614 pmol/l were 3.4 and 3.0%, respectively. Equivalent inter-assay CVs were 1.9 and 10.0%, respectively.

Data for BMI standard deviation score (SDS) and HbA_1c were normally distributed (as confirmed by the One-Sample Kolmogorov-Smirnov Test). Each case was matched against two control subjects. All analyses compared the mean data from case subjects (weight, height, insulin dose, C-peptide level, and HbA_1c) against the mean data from the two matched control subjects. A general linear ANOVA model was used for comparison of data between the groups and data within case subjects. For each individual, the proportion of the observed period spent on a particular insulin regimen was expressed as a percentage of the total observation period. These data were then compared between the groups using the χ² test. An ANOVA model was used to determine the influence of celiac disease on both HbA_1c and BMI SDS, allowing for factors such as insulin dose, insulin regimen, and time. Time was expressed as months relative to onset of celiac disease. Data are presented as mean values ± SEM unless otherwise stated, and a P value < 0.05 was considered significant.

Height, weight, and BMI SDS (measured as kilograms per meter squared) scores were calculated using reference data based on the British 1990 Growth Reference and Cole’s LMS method supplied by the Childhood Growth Foundation (21). SPSS version 10 was used for all statistical analysis and significance.

Cohort characteristics are described in Table 1. In case subjects compared with control subjects, mean BMI SDS was significantly lower (−1.2 ± 0.1 vs. −0.1 ± 0.1, P = 0.005), as was mean weight SDS (−0.7 ± 0.3 vs. 0.5 ± 0.3, P = 0.002) (Table 1). There was no difference between the groups in mean height SDS (0.1 ± 0.2 vs. 0.7 ± 0.3, P = 0.19) or mean C-peptide level (72.8 ± 23 vs. 76.1 ± 13 pmol/l, P = 0.32) (Table 1). For case subjects diagnosed at first antibody testing (i.e., within 1 year of onset of diabetes) compared with case subjects diagnosed in >1 year of onset of diabetes, no difference was found in BMI SDS.

Cohort characteristics at diagnosis of celiac disease are described in Table 1. The mean age of diagnosis of celiac disease was 11.2 years (range 2.2–17.3). The mean duration of diabetes at diagnosis of celiac disease was 3.8 years (range 0.9–7.2). Of the 11 case subjects, 3 were detected on their first annual screening test. The mean number of annual screening tests was four per subject, range 1–7. In the period between diagnosis of type 1 diabetes and the diagnosis of celiac disease, mean BMI SDS remained significantly lower in case subjects compared with control subjects (−0.2 ± 0.2 vs. 0.7 ± 0.2, P = 0.015) (Table 2) (Fig. 1), as was mean HbA_1c (8.9 ± 0.3 vs. 9.8 ± 0.3, P = 0.002) (Table 2) (Fig. 2). Clinical mean HbA_1c for this age group was 9.3%. There was no difference in daily insulin dose between the groups (0.8 ± 0.1 vs. 0.8 ± 0.1 units · kg^–1 · day^–1, P = 0.54 for cd⁺ vs. cd⁻), but case subjects received less-intensive insulin regimens compared with the cd⁻ group (as expressed by the proportion of the total observation period spent on a particular regimen, χ² = 7.6, P = 0.019) (Table 2). There was no difference in mean C-peptide levels (13.2 ± 2.0 vs. 26.5 ± 5.6 pmol/l, P = 0.22 for cd⁺ vs. cd⁻).

Of the 11 case subjects, 6 agreed to a repeat small bowel biopsy on a GFD that was consistent with the diagnosis of celiac disease and successful treatment. All subjects reverted to antibody-negative status within 3–6 months of starting a GFD and have remained negative to date. Symptoms resolved in the single subject who was symptomatic at diagnosis of celiac disease. The difference in mean BMI SDS between case subjects and control subjects at diagnosis of celiac disease was reversed over 12 months after the start of a GFD (1.1 ± 0.13 vs. 1.0 ± 0.1, P = 0.11) (Table 2). Thereafter, there was no significant difference in mean BMI SDS between the groups. To date, the control subjects remain antibody negative on annual screening.

For case subjects, mean HbA_1c levels were lower for the period on a GFD compared with before diagnosis of celiac disease (8.9 ± 0.1 vs. 8.3 ± 0.1, P = 0.002). Furthermore, on a GFD, mean HbA_1c levels remained significantly lower in case subjects compared with control subjects (8.3 ± 0.2 vs. 10.0 ± 0.2%, P = 0.02) (Table 2).

In both groups, insulin requirements increased compared with before a GFD was started (0.8 ± 0.1 vs. 1.0 ± 0.04, P < 0.001 for pre- vs. post-GFD), but there was no difference in daily insulin dose (units · kg⁻¹ · day⁻¹) between the groups (1.0 ± 0.04 vs. 1.0 ± 0.03, P = 0.33 for cd⁺ vs. cd⁻). In the follow-up period after a GFD was started, there was no statistical difference in the proportion of time spent on a particular insulin regimen between the groups (P = 0.20) (Table 2).

From diagnosis of diabetes to diagnosis of celiac disease, in using a general factorial linear model, celiac disease was associated with lower BMI SDS (F = 32.6, P < 0.0001) and a lower HbA_1c (F = 4.2, P = 0.04) across time (months relative to onset of celiac disease), independent of other factors, namely insulin dose and insulin regimen. Similarly, in the period on the GFD, celiac disease was associated with lower HbA_1c (F = 5.1, P = 0.027), independent of BMI SDS, insulin dose, and insulin regimen.

We report results from a longitudinal case-controlled study detailing the impact of celiac disease on children with type 1 diabetes before and after intervention with a GFD. Untreated celiac disease was associated with lower BMI SDS and lower HbA_1c. Introduction of a GFD led to recovery of BMI SDS and further improvement in HbA_1c. In contrast to control subjects, in case subjects there was no deterioration in glycemic control during the pubertal years.

Symptomatic celiac disease shows a dramatic response to intervention with a GFD (15,22); however, serological testing now permits an earlier diagnosis of celiac disease (23, 24). Prevalence rates in the region of 1.7–10% have been reported in the pediatric diabetes clinic setting (2–5), with many case subjects being asymptomatic or having atypical features only recognized in retrospect (7).

Previous studies investigating the effect of celiac disease on type 1 diabetes and the benefit of treatment with GFD have been cross-sectional or confounded by lack of control subjects (5,11, 14,17,20–22), relied on inadequate longitudinal follow-up data (10–13,19, 25), or were disrupted by poor compliance to a GFD (13,14); thus, data are conflicting. A question therefore remains as to whether it is worthwhile to screen for silent celiac disease and whether subjects benefit from a GFD in terms of metabolic control and growth.

At diagnosis of diabetes, our data show lower BMI SDS in case subjects compared with control subjects, and as far as we know, this has not been previously described. Although we have no information on EMA/AGA status at diagnosis of diabetes, this may indicate the presence of silent celiac disease. This may be further evidence of a possible common immunopathology in genetically susceptible individuals; however, the pathogenetic mechanisms remain unexplained (23,26).

From diagnosis of type 1 diabetes to diagnosis of celiac disease, lower BMI SDS in our case subjects has again been inadequately described in previous studies. Some report short stature with no differences in weight (4,5,11, 16), whereas others have shown no differences in height or weight (15–17,19, 25). During the same period, HbA_1c levels were lower in case subjects despite no difference in C-peptide levels. Both these features may be consistent with the effects of villous atrophy secondary to celiac disease and subsequent carbohydrate malabsorption.

Once established on a GFD, the difference in BMI SDS took 12 months to disappear, presumably reflecting slow recovery of small bowel morphology. Others have previously shown no difference in growth parameters regardless of dietary compliance (20). Within case subjects, the start of a GFD resulted in further significant improvement in mean HbA_1c levels compared with pre-GFD, confirming findings of some studies (24,27) but not others (13,14, 19,25). Furthermore, mean HbA_1c levels were also lower when compared with control subjects. The period of follow-up on a GFD was in the pubertal age range, when even well-motivated intensively treated patients invariably demonstrate worsening glycemic control compared with prepuberty, which in part is related to increased insulin resistance (28). As expected, this occurred in the control group, with a gradual rise in HbA_1c. In contrast, those with celiac disease did not show any worsening of glycemic control, suggesting that intervention with a GFD is associated with improvement (or lack of deterioration) in glycemic control compared with control subjects. We found no difference in C-peptide levels, insulin dose, or regimen to account for these findings, which may be due to small sample size and increased dietetic input received by case subjects. An alternative explanation is that a GFD may influence insulin sensitivity. The glycemic index of gluten-free and gluten-containing foods is similar (29). However, data suggest that the type of carbohydrate may be important in influencing insulin sensitivity (30). The gain in BMI SDS and lack of deterioration in HbA_1c in case subjects during puberty while on a GFD, with no difference in insulin dose between the groups, would be compatible with this explanation. No difference in HbA_1c between the groups was found at the end of the follow-up period, and this may signify that puberty is complete and differences in insulin sensitivity are less pronounced.

This study is limited by small sample size, yet detailed follow-up is longer than previously published data, with case subjects being tightly matched to control subjects. Mean HbA_1c levels in the control group were higher than the clinic mean, which may reflect unmatched factors that impact glycemic control. In addition, initial data are potentially skewed by the presence of three case subjects diagnosed within 1 year of diagnosis of diabetes. However, comparisons of these against the remaining case subjects revealed no significant differences and therefore our initial observations remain valid. Only six subjects were re-biopsied on a GFD to ensure compliance with diet, although all case subjects maintained a negative antibody status throughout the study period.

A recent review by Holmes (9) recommended routine antibody screening on the basis of the high prevalence of silent celiac disease in type 1 diabetic subjects and the availability of a cost-efficient method for screening. The long-term sequelae of untreated celiac disease, such as gastrointestinal malignancy, lymphoma, infertility, and osteoporosis, would also suggest long-term benefit from a GFD (31–33). We have demonstrated in a longitudinal case-controlled study in children with type 1 diabetes that silent celiac disease results in poor weight gain and lower HbA_1c independent of daily insulin dose and regimen. Intervention with a GFD causes BMI SDS to recover and HbA_1c to improve further within case subjects compared with before the GFD. In addition, case subjects in contrast to control subjects did not demonstrate the deterioration in glycemic control that invariably occurs during puberty. These apparent clinical benefits need confirming by larger, similarly designed studies before routine antibody screening is universally adopted.

We gratefully acknowledge assistance from S. Gould (consultant pediatric pathologist), P. O’Sullivan (consultant pediatric gastroenterologist), S. Strang, A. Watts, and D. Harris at the John Radcliffe Hospital, Oxford.