Celiac Autoimmunity Is Associated With Lower Blood Pressure and Renal Risk in Type 1 Diabetes

Abstract

Context

Though the long-term consequences of celiac disease (CD) in type 1 diabetes are unclear, CD has been associated with increased prevalence of end-stage renal disease (ESRD) independent of type 1 diabetes.

Objective

We evaluated whether celiac autoimmunity is related to the cumulative incidence of microalbuminuria [albumin excretion rate (AER) 20 to 200 µg/min], macroalbuminuria (AER >200 µg/min), and ESRD.

Design, Patients, and Methods

In the prospective follow-up of the Pittsburgh Epidemiology of Diabetes Complications study of childhood-onset type 1 diabetes, 618 participants were screened for tissue transglutaminase (tTG) antibodies with a clinical assay. Nephropathy outcomes were determined at 25 years of diabetes duration.

Results

Overall, the 33 subjects (5.3%) with strongly positive tTG levels (≥3 times the upper limit of normal) or a reported clinical history of CD had lower baseline blood pressure and lipid values. At 25 years of diabetes duration, a lower cumulative incidence of macroalbuminuria in strongly positive subjects compared with those with negative serology (3.6% vs 30.0%; P = 0.003) remained significant after adjustment for age, HbA_1c, lipid measures, and blood pressure (adjusted P = 0.004). No considerable differences between these subjects and tTG-negative groups were found for microalbuminuria (40.0% vs 57.1%) or ESRD (0 vs 4.1%).

Conclusions

These findings show that strongly positive celiac autoimmunity status in individuals with childhood-onset type 1 diabetes is associated with lower baseline blood pressure and cholesterol measurements as well as lower macroalbuminuria risk after 25 years of type 1 diabetes duration with no increase in the risk of microalbuminuria or ESRD.

Celiac disease (CD) is an autoimmune enteropathy resulting from both genetic and environmental factors. It is unique in that its environmental trigger is known and, with gluten exclusion, can be successfully treated. The prevalence of CD in type 1 diabetes is between 1.6% and 9.7% worldwide (1), much higher than the general US population prevalence of 0.71% (2). With the advent of serologic assays as a noninvasive screening tool, those with asymptomatic or atypical, extraintestinal manifestations are being diagnosed more commonly, making the classic presentation of diarrhea and weight loss less common (3–5).

Routine screening for CD in patients who are asymptomatic with type 1 diabetes remains controversial (6–9), as the long-term consequences of undiagnosed CD in individuals with type 1 diabetes are unclear, and gluten restriction is difficult (10), particularly in combination with dietary restrictions for diabetes. Diagnosis of CD in adults with type 1 diabetes in clinical practice ranges from a case-finding approach based on clinical symptoms to routine screening of all patients. In particular, a recent review of ∼50 published articles noted that although CD screening at type 1 diabetes diagnosis and within 2 and 5 years thereafter was a clear evidence-based recommendation, because 55% of cases were diagnosed within 2 years and 79% diagnosed within 5 years of diabetes duration, they noted more research was needed regarding CD screening beyond 5 years of diabetes duration (1).

In the general population, CD is reported to be associated with an increased risk of IgA nephropathy (11) and end-stage renal disease (ESRD) independent of type 1 diabetes (12). However, studies in individuals with type 1 diabetes and CD are conflicting, and the mode of renal disease assessment varies. There are reports of increased stage III chronic kidney disease with only borderline nonsignificant changes in albumin excretion rate (AER) (13), no relationship between CD and presence of microalbuminuria (14) or persistent macroalbuminuria (15), and increased risk of nephropathy defined by International Classification of Diseases, Ninth Revision, codes only after >10 years of CD (16). Using the prospective observational Pittsburgh Epidemiology of Diabetes Complications Study (EDC) cohort who have had detailed assessments of renal function, we evaluated whether celiac autoimmunity assessed by standard clinical assays of tissue transglutaminase (tTG) antibodies or known history of CD augments the development of microalbuminuria, macroalbuminuria, and ESRD over 25 years of diabetes duration in patients with childhood-onset type 1 diabetes. Because the tTG assay was not available when our cohort was originally enrolled (1986 to 1988), we wanted to determine if knowledge of celiac status in the course of diabetes, either by antibody or clinical diagnosis status, is associated with any adverse or protective effects in type 1 diabetes renal outcomes.

Subjects and Methods

Study population

The EDC study is a prospective type 1 diabetes cohort that comprises those who were diagnosed with childhood-onset type 1 diabetes, at an age younger than 17 years, at Children’s Hospital of Pittsburgh (CHP) between 1 January 1950 and 31 May 1980 and living within 100 miles or 2.5 hours’ drive from Pittsburgh. The cohort has been described in detail elsewhere (17, 18), and the CHP registry has been shown to be epidemiologically representative of community-based type 1 diabetes (19). Informed consent was obtained from all patients with institutional review board approval.

Of 1124 eligible participants, 145 (13%) had died before baseline (1986 to 1988), and 191 (17%) chose not to participate. Of the 788 participants (70%), 130 (16%) provided survey information only, and 658 patients had both a baseline examination and a survey. Baseline examinations were conducted 1986 through 1988 and used as baseline characteristics for this study (Fig. 1). Participants were examined biennially for the first 10 years of the study and then again at 18 years and 25 years. Surveys have been conducted biennially since the start of the study. These subjects were cared for prior to the availability of serological assays to evaluate celiac autoimmunity.

Celiac autoimmunity definitions

A stored serum sample from each participant was tested for tTG antibodies at either the 25-year exam (2011 to 2014) or from the latest previous exam based on sample availability. Serum was stored at −80°C, and to represent assays used in clinical practice, samples were analyzed in a clinical laboratory for tTG IgA measured by the Quanta Lite h-tTG IgA ELISA (Inova Diagnostics, San Diego, CA) with sensitivity and specificity being 92.7% and 91.6%, respectively. Coefficient of intra-assay and interassay variation ranged between 2.3% to 4.8% and 1.8% to 5.8%, respectively, using interassay control samples determined as negative, weakly positive, and strongly positive. An IgA level was measured in the same serum sample to detect IgA deficiency, and, if in the deficient range, the same serum sample or, if not available, a serum specimen from a cycle closest in time to the original specimen was sent to the laboratory for tTG IgG measurements using a Quanta Lite R h-tTG IgG ELISA (Inova Diagnostics). Sensitivity and specificity of this assay were 78.8% and 74.3%, respectively. Coefficient of intra-assay and interassay variation ranged between 3.5% and 5.7% and 6.9% and 11.0%, respectively.

Positive celiac autoimmunity was defined as exceeding the upper limit of normal (ULN) either as a tTG IgA ≥20 U/mL or a tTG IgG ≥6 units. Positive autoimmunity was further classified into strongly positive autoantibodies defined as values ≥3 times ULN (tTG IgA ≥60 units or tTG IgG ≥18 units), as lower levels are reportedly less specific for clinically noteworthy CD (9). Known CD was classified clinically by patient history through either biennial questionnaires or self-report at clinical exams. Participants who self-reported a previous clinical diagnosis of CD with undetectable antibodies (n = 4) were also classified with the strongly positive autoimmunity group regardless of antibody status at the time of testing.

Ascertainment of diabetes complications

The cumulative incidence of renal complications was assessed up to 25 years of type 1 diabetes duration for each subject. Participants provided three timed urine samples (24-hour, overnight, and 4-hour collections obtained over a 2-week period) at each clinical examination. Urinary albumin was measured using immunonephelometry (20), and AER was calculated for each sample. Microalbuminuria was defined as the first instance of AER ≥20 μg/min (also known as moderately increased albuminuria), and macroalbuminuria was the first instance of AER >200 μg/min (also known as severely increased albuminuria) (21) confirmed in at least two of the three timed urine collections. ESRD was the first instance of self-reported renal transplantation or dialysis. Glomerular filtration rate was estimated using the Chronic Kidney Disease Epidemiology Collaboration creatinine equation (22).

Risk factors and clinical assessments

At the EDC exam, height was measured using a stadiometer, and weight was measured on a calibrated balance beam scale. Standardized sitting blood pressures and heart rate were measured after a 5-minute rest period. Hypertension measured with a random zero sphygmomanometer was defined as systolic blood pressure (SBP) ≥140 mm Hg or diastolic blood pressure (DBP) ≥90 mm Hg or the reported use of medications for blood pressure control. For the first 18 months of the study, stable HbA_1c was measured at baseline by ion-exchange chromatography (Isolab, Akron, OH) and, for the remainder of the baseline exams, automated HPLC (Diamat; Bio-Rad, Hercules, CA) was performed. The two assays were highly correlated [r = 0.95; Diamat (HbA_1c) = −0.18 + 1.00 (Isolab HbA_1c)]. Values were converted to Diabetes Control and Complications Trial–aligned HbA_1c using a regression equation derived from duplicate assays [Diabetes Control and Complications Trial HbA_1c = 0.14 + 0.83 (EDC HbA_1c)] (23). Total, high-density lipoprotein (HDL), and low-density lipoprotein (LDL) cholesterol and triglycerides were determined as previously described (24–27). An ever-smoker was one who smoked ≥100 cigarettes over his or her lifetime. All participants completed a questionnaire regarding medical history before each exam.

Data analysis

The cumulative incidences of complications of those with weakly positive and strongly positive celiac autoantibodies were compared with those negative for celiac autoantibodies. Participants who reported a complication at the study baseline exam and had diabetes duration ≤25 years were classified as an incident case. For the 25-year cumulative incidence, a participant was classified as an incident case if he or she experienced the complication before or at 25 years’ duration of diabetes. Participants who had diabetes duration of >25 years at study baseline and were found to have evidence of nephropathy (n = 38 with microalbuminuria and n = 48 with macroalbuminuria) were excluded from the 25-year incidence analysis of that particular complication, as the date of complication incidence was unknown, and we could not confirm whether the complication occurred before 25 years. However, these subjects were included in the analyses for the absence of a more severe renal complication at 25 years. For example, if at EDC baseline, a participant had diabetes duration of 30 years and macroalbuminuria, they did not have ESRD at 25 years, but status for microalbuminuria and macroalbuminuria were considered unknown because the timing of albuminuria could not be fully assessed. A total of 17 participants with ESRD at study entry and diabetes duration of >25 years were excluded from the analysis entirely because their renal status at 25 years was unknown.

Data were summarized as means and SDs or medians with interquartile ranges for continuous variables and as frequencies and percentages for categorical variables. The distributions were assessed for normality. Two-sample t tests were used to compare risk factors between the negative and weakly positive celiac autoantibody group and, separately, between the negative and strongly positive groups. Wilcoxon rank-sum test was used if the risk factors were not normally distributed. For categorical variables, a χ² analysis was used to compare frequencies. Fisher exact test was performed if expected counts were <5. A P value of <0.05 was considered statistically significant. Logistic regression was used to assess the association of 25-year complication status with celiac autoimmunity, adjusting for blood pressure and lipid profile. All statistical analyses were performed using SPSS version 24 (IBM Corp, Armonk, NY).

Results

Celiac autoimmunity status

Of 641 participants who provided serum samples, 618 had specimens available for measurement of celiac serology and were classified by autoimmunity status (Fig. 2). Normal IgA levels were found in 582 (94.2%) subjects, of whom 49 (8.4%) had positive tTG IgA autoantibodies. Of this group, based on tTG IgA levels, 22 (45.0%) were weakly positive with tTG levels one to three times the ULN, including 2 with known CD, and 27 (55.1%) were strongly positive with tTG levels >3 times the ULN, including 6 with known CD. Of the 533 participants with normal IgA levels and negative tTG IgA autoantibodies, 2 reported a prior diagnosis of CD. There were 36 (5.7%) participants with IgA deficiency, of whom 13 (36.1%) had positive tTG IgG autoantibodies, with 2 (9.0%) of these 13 being strongly positive with no reported cases of clinical CD. Therefore, a total of 64 participants (10.4%) had evidence of celiac autoimmunity either by serology or clinical history, with 33 (5.3%) of the total participants having strongly positivity autoimmunity or reported clinical CD.

Baseline characteristics

The characteristics at the baseline EDC examination (1986 to 1988) by celiac autoimmunity groups are detailed in Table 1. There were no considerable differences in age of diabetes onset or type 1 diabetes duration, race, sex, HbA_1c, or body mass index (BMI) measured at the baseline examination between those with negative and positive celiac status. Men with weakly positive tTG were shorter than those with negative tTG (P = 0.008), but no other significant differences in height were noted between groups. Compared with the negative group (Table 1), those strongly positive levels had lower baseline SBP (P = 0.001), DBP (P = 0.014), and prevalence of a hypertension diagnosis (P = 0.045). The difference in blood pressure can also be seen in Fig. 3, in which 45.5% of strongly positive celiac subjects had baseline SBP levels in the lowest quintile (P = 0.005 compared with negative group). No differences existed in the groups regarding antihypertensive therapy, estimated glomerular filtration rate (eGFR), or current or previous history of smoking. Lipid profiles did not differ between those with weakly positive and negative celiac status. Subjects classified as strongly positive had lower total cholesterol (P = 0.001), LDL (P = 0.030), HDL (P = 0.035), and triglycerides (P = 0.001) than those with negative celiac status. Of participants with ≥25 years diabetes duration at baseline, 59 had no evidence of renal disease (none were antibody positive), 38 had microalbuminuria (1 strongly antibody positive and 4 weakly antibody positive), and 52 had macroalbuminuria (1 strongly antibody positive).

Renal outcomes

Clinically notable celiac autoimmunity (high-titer tTG or clinical CD) was associated with a 25% lower prevalence of microalbuminuria compared with weak or negative celiac autoimmunity groups. The strongly positive group had significantly lower incidence of macroalbuminuria (3.6%) than seen in those with negative serology (30.0%; P = 0.003). This pattern persisted after excluding the two subjects with a history of CD and negative antibodies, one of whom had only microalbuminuria, and the other had no evidence of nephropathy. There were no marked differences in the cumulative incidence of microalbuminuria or ESRD by 25 years of diabetes duration by celiac autoimmunity groups (Fig. 4). The lower incidence of macroalbuminuria in strong celiac autoimmunity remained statistically significant after adjusting for age, HbA_1c, blood pressure, and lipid measures (adjusted P = 0.004). A total of six participants were censored prior to 25 years’ duration (two deaths and four lost to follow-up).

When compared by autoimmunity status, there was no difference in eGFR of <60 mL/min/1.73 m² at 25 years in those strongly (3.1%) or weakly (3.6%) positive compared with those negative (9.5%) for celiac autoimmunity. There was an incremental pattern of smaller average annual and percent eGFR decline in the strongly positive group (−1.25 mL/min/1.73 m²; −1.0%) compared with the weakly positive (−1.31 mL/min/1.73 m²; −1.9%) and negative (−1.91 mL/min/1.73 m²; −2.1) groups, although these trends did not reach statistical significance (P = 0.37 and 0.19, respectively).

There were no differences in HLA DR3 or DR4 positivity between nephropathy groups, with 89.8% of the entire group DR3 or DR4 positive, and no difference by DR3/DRX (where X is not DR4) status (microalbuminuria, 27.4% vs 22.8%; macroalbuminuria, 27.8% vs 23.4%; and ESRD, 35.0% vs 24.3%). By celiac status, 43.8%% of those strongly positive, 36.7% of those weakly positive, and 24.2% of those negative were positive for HLA DR3 without HLA DR4 (P = 0.019).

Of interest, no gastrointestinal malignancies were reported.

Discussion

These results suggest that celiac autoimmunity in patients with type 1 diabetes does not increase the risk of renal disease and, in this study, is associated with a lower incidence of macroalbuminuria. Although the mechanism remains unclear, celiac autoimmunity is associated with lower blood pressure and lipid measures and may be related to some malabsorption. In particular, after controlling for the lower blood pressure and lipid profile, the lower incidence of nephropathy remained considerable in those with type 1 diabetes with either clinically diagnosed CD or strongly positive tTG antibodies. These findings of lower blood pressure and lipid measures are supported by another study comparing adults with diagnosed CD to age-matched and sex-matched controls. In that study, West et al. (28) demonstrated a lower prevalence of the diagnoses of hypertension and hypercholesterolemia in 3790 patients with CD compared with 17,925 of the general population. However, a much smaller case-control study comparing 31 cases with type 1 diabetes and CD to 46 with type 1 diabetes alone found no differences in blood pressure or use of blood pressure–lowering drugs (14). Differences in lipid profile, including a low HDL, have previously been reported in CD and speculated to result from malabsorption, decreased circulating lipoproteins, and decreased secretion of apolipoprotein from the intestinal mucosa (29).

Our study specifically studied celiac autoimmunity over a duration of diabetes of 25 years and found a reduced risk of nephropathy in those with CD or high-titer tTG. The findings of a lower incidence of macroalbuminuria and no difference in stage III kidney disease reported in this study differ somewhat from those of a prior study in type 1 diabetes (13). In that report, previously undetected CD was associated with a higher prevalence of stage III kidney disease at the time of CD diagnosis compared with controls matched for age, sex, weight, insulin dose, and duration of diabetes. However, those patients with CD also had worse glycemic control, with an HbA_1c of 8.2% compared with 7.5%, which could have contributed to their higher prevalence of overall microvascular complications (13). A population-based study of biopsy-confirmed CD and type 1 diabetes in the Swedish population found an increased risk of chronic renal disease using International Classification of Diseases, Ninth Revision codes, only after >10 years of CD in patients with a median age at the end of the study of 21 years. The contribution of duration of type 1 diabetes and clinical measures of HbA_1c and cholesterol to renal disease was not reported (16). In contrast, two other studies found no notable difference with regard to diabetic renal status and CD (14, 15). HbA_1c was similar in the celiac and nonceliac cases in both of these reports, as it is in our analyses. One of these two studies also found a lower prevalence of retinopathy in the CD group, whereas no differences in blood pressure or use of blood pressure–lowering drugs existed (14). This lower prevalence of retinopathy was hypothesized to be due to the possibility of a favorable effect of a gluten-free diet on the development of vascular complications, but this particular retrospective study was unable to biochemically confirm the efficacy of the self-reported gluten-free diet.

A study in children with type 1 diabetes reported lower urinary albumin-to-creatinine ratios in those with a biopsy-confirmed CD compared with those without CD and that those with CD also had a slower progression in albuminuria over 5 years of follow-up (30). Unlike our study, these participants were demonstrated to be compliant with a gluten-free diet by measurement of tTG IgA antibodies at the start of the study. There were no differences in blood pressure, HbA_1c, or BMI between these groups of children. The authors concluded that a gluten-free diet might have renoprotective effects (30). Another report found potential renoprotective effects of a gluten-free diet in children with type 1 diabetes, biopsy-proven CD, and negative tTG antibodies who had lower urinary albumin-to-creatinine ratios and lower levels of circulating advanced glycation end products, which are less commonly found in gluten-free foods, as this diet is low in high-temperature processed foods and flours (31). Our data would suggest a protective effect of high-titer celiac antibody positivity, even in the absence of diagnosed CD or confirmed gluten restriction. Adult patients with type 1 diabetes with CD have also been shown to have lower levels of coagulation factors, suggesting attenuation of the prothrombotic state of type 1 diabetes by CD (32).

The strengths of the analyses in this report are the size and representativeness of the original cohort, long-term follow-up, careful and standard ascertainment of baseline characteristics, clinical measures, and complications prospectively, and the ability to evaluate the effects of CD autoimmunity per se as most were minimally or untreated with a gluten-free diet when examined retrospectively based on their celiac antibody titers. Studies of the EDC population have shown that even though participants were diagnosed within a wide time frame during which many advances in the diagnosis and treatment of diabetic nephropathy have occurred, little evidence exists for declines in cumulative incidence of renal outcomes in recent cohorts except for ESRD (33). A limitation of this analysis is the lack of information on the date of complication incidence in those cases who were prevalent cases at baseline. Thus, we were only able to classify these cases as positive or negative by 25 years of diabetes duration and are unable to perform time-to-event analyses (e.g., Cox regression).

We were unable to ascertain the onset and duration of CD, and the majority of subjects with even strongly positive antibody testing had no known CD diagnosis at the time of testing. This is despite the long duration of diabetes in these participants, which averaged 17 years at the baseline examination (1986 to 1988) with probable long-term CD autoimmunity, which usually develops in childhood. This concept is confirmed by recent data suggesting that CD autoantibodies appear at or soon after diagnosis of type 1 diabetes, persist as long as CD is untreated, and few appear to acquire these antibodies after puberty (34, 35). It can be argued that there was a lack gastroenterologic gold-standard biopsy confirmation of CD in all cases. However, it has been demonstrated that tTG antibodies ≥3 times the ULN correlate with greater likelihood ratios for CD based on intestinal biopsies (36). Our study was primarily designed to explore the role of routinely screened CD autoimmunity assessment, as a marker of intestinal inflammation, on renal outcomes. In support of the significance of CD autoantibodies, a recent study in Finland (37) showed that patients with celiac antibodies detected during type 1 diabetes surveillance had similar clinical and histological results compared with those with CD diagnosed on a clinical basis. Additionally, the relative hypocholesterolemia and lower blood pressure in strongly positive subjects compared with negative subjects is consistent with patterns previously observed in CD presumably due to the associated malabsorption.

An additional limitation is lack of ascertainment of duration and compliance of a gluten-free diet in those with known CD because our assessment of CD was retrospective. There were only 2 subjects with clinically diagnosed CD with negative tTG levels likely reflecting excellent dietary compliance, whereas 8 of the 10 subjects with known CD had positive tTG titers, likely reflecting lower compliance and thus a continued effect of the CD. It thus seems unlikely that the effects of a gluten-free diet in our population are contributing to the lower incidence of nephropathy. The lower sensitivity and specificity of the tTG IgG assay reflect the known reduced capacity of this clinical test to make an accurate diagnosis in patients who are IgA-deficient, and only two subjects met the strongly positive criteria by this assay, so CD may have been missed in some participants.

In conclusion, we examined whether evidence of CD or celiac autoimmunity, representing either undiagnosed or uncontrolled CD, had an effect on diabetic renal outcomes after 25 years of type 1 diabetes duration. These results suggest that celiac status, rather than exacerbating nephropathy, has a protective effect on blood pressure, lipid profile, and the development of macroalbuminuria. The exact mechanism of such effects remains unclear but may be related to mild malabsorption. Recognizing this association may lead to a better understanding of the underlying pathophysiology and risk for development of macroalbuminuria. This protective effect needs to be confirmed for other type 1 diabetes morbidities and weighed against CD-related complications to assess the beneficial effects of the burdensome total restriction of gluten in subjects with no clinical features of CD, especially those with low-titer tTG. Based on our findings, relying on antibody screening alone for CD will not likely influence the outcome of nephropathy, whereas evaluation of clinical symptoms remains important for appropriate identification and treatment of CD.

Abbreviations:

Abbreviations:
 
• AER

  albumin excretion rate

 
• BMI

  body mass index

 
• CD

  celiac disease

 
• CHP

  Children’s Hospital of Pittsburgh

 
• DBP

  diastolic blood pressure

 
• EDC

  Epidemiology of Diabetes Complications

 
• eGFR

  estimated glomerular filtration rate

 
• ESRD

  end-stage renal disease

 
• HDL

  high-density lipoprotein

 
• LDL

  low-density lipoprotein

 
• SBP

  systolic blood pressure

 
• tTG

  tissue transglutaminase

 
• ULN

  upper limit of normal

Acknowledgments

We thank all of the study participants who volunteered time and the CHP laboratory and EDC staff.

Financial Support: This research was supported by National Institutes of Health Grant DK34818 and the Rossi Memorial Fund.

Author Contributions: K.V.W. was responsible for data review, analyses, and writing the final version of the manuscript. C.L.C. was responsible for data compilation, initial analyses, and the first draft. R.G.M. was responsible for data set compilation and analyses. V.C.A. provided data analysis, writing, and editing of the manuscript. I.L. contributed to interpreting the data and revising the manuscript. Y.H. provided data analysis. D.J.B. was responsible for the concept, assistance with design, handling of laboratory assessments of rTG antibodies, and manuscript editing. T.J.O. was responsible for study design, data collection, and revision of the manuscript.

Disclosure Summary: The authors have nothing to disclose.

References