Prognosis of unrecognized coeliac disease as regards mortality: A population-based cohort study

Abstract

Background and aim. Clinically diagnosed coeliac disease patients carry an increased risk of mortality. As coeliac disease is markedly underdiagnosed, we aimed to quantify the risk of mortality in subjects with unrecognized and thus untreated coeliac disease.

Method. Blood samples from 6,987 Finnish adults were drawn in 1978–80, and sera were tested for immunoglobulin A (IgA)-class tissue transglutaminase antibodies (Eu-tTG) in 2001. Positive sera were further analysed for endomysial (EMA) and tissue transglutaminase antibodies by another test (Celikey tTG). EMA- and Celikey tTG-positive cases were compared to negatives as regards mortality in up to 28 years of surveillance, yielding a total follow-up of 147,646 person years. Dates and causes of death were extracted from the nation-wide database.

Results. Altogether 74 (1.1%) of the participants were EMA- and 204 (2.9%) Celikey tTG-positive. The age- and sex-adjusted relative risk of overall mortality was not increased in either EMA (0.78, 95% CI 0.52–1.18) or Celikey tTG (1.19, 95% CI 0.99–1.42) -positive subjects. However, antibody-positive cases evinced a tendency to die from lymphoma, stroke, and diseases of the respiratory system.

Conclusions. The prognosis of unrecognized coeliac disease was good as regards overall mortality, which does not support screening of asymptomatic coeliac disease cases.

• Antibody
• coeliac disease
• epidemiology
• mortality

Introduction

Coeliac disease is a chronic autoimmune-like disorder with intestinal and extraintestinal manifestations induced by wheat gluten and related proteins of rye and barley. Conceptions of the epidemiology of coeliac disease have changed substantially over time. It is currently considered a global prevalent disorder increasing over time and affecting up to 1%–2% of the Western population [1–5]. Diarrhoea and malabsorption as clinical manifestations are nowadays more rarely seen, and up to 70%–90% of the coeliac disease population in Western countries remain unrecognized due to the absence or the atypical nature of symptoms [1–5]. Prognostic studies on undiagnosed and thus untreated cases are scant, and debate thus continues as to whether health care professionals should rigorously seek out these cases even by population mass-screening [6–11].

Diagnosed clinically detected coeliac disease cases carry a 1.3–3.8-fold increased risk of mortality mainly attributable to malignancies, and the risk seems to decrease on a gluten-free diet [12–20]. However, there have been only two previous attempts to assess the association between unrecognized coeliac disease and mortality [21], [22]. Due to the relatively short follow-up times and modest number of antibody-positive cases in the studies in question, the results remain inconclusive.

Key messages

• The prognosis of undetected coeliac disease is good as regards overall mortality.

• Coeliac antibody-positive undetected cases evinced an additional tendency to die from lymphoma, stroke, and diseases of the respiratory system.

We followed a large Finnish population-based adult-representative cohort of 6,987 people collected in 1978–80, focusing on mortality and aiming to establish whether unrecognized coeliac disease individuals carry an increased risk of mortality during a surveillance of up to 28 years. Sequential analysis of sera from the participants by tissue transglutaminase and endomysial antibody tests with reported high validity [23], [24] enabled us to define coeliac autoantibody-positive cases most likely representing unrecognized coeliac disease. We availed ourselves of the causes of death register of Statistics Finland and could compare the overall and cause-specific mortality between antibody-positive and -negative subjects throughout the follow-up period.

Material and methods

Study population

The Mini-Finland Health Survey carried out in 1978–80 provided a basis for the current population-based follow-up study between autoantibody-positive unrecognized coeliac disease and mortality. Detailed information on the design and base-line results of the primary study has been published elsewhere [25], [26]. In brief, a nationally representative sample of 8,000 persons was drawn from the population aged 30–99 years according to a stratified two-stage cluster-sampling model planned by Statistics Finland. The second stage of sampling was implemented in 40 areas in different parts of the country. The participants attended a health examination which included interviews, questionnaires, drawing of blood samples, and a clinical examination by a physician.

Abbreviations

Table

Celikey	a test for immunoglobulin A (IgA)-class
EMA	a test for IgA endomysial antibodies
Eu-tTG	a test for IgA-class tissue transglutaminase antibodies (Eu-tTG® umana IgA, Eurospital S.p.A., Trieste, Italy)
ICD	International Classification of Diseases
IgA	Immunoglobulin
tTG	tissue transglutaminase antibodies (Celikey®, Phadia, Freiburg, Germany)

The participation rate was 90% (n =7,217), and sera from 6,990 individuals were still available for the purposes of this study in 2001 (Figure 1). Three coeliac disease or dermatitis herpetiformis cases previously diagnosed and treated were excluded from the analysis at the beginning of the follow-up, yielding a total of 6,987 participants (3,766 females, mean age 51 years, age range 30–95) for this study (Figure 1). Information on age, sex, education, body mass index, alcohol consumption, smoking, hypertension, serum cholesterol, low-density lipoprotein, high-density lipoprotein, triglycerides, diabetes, coronary heart disease, stroke, and cancer were extracted for purposes of adjustment. The definitions and measurement of potential confounding and effect-modifying factors have been described in greater detail elsewhere [25], [26]. All participants gave informed consent. The Ethical Committee of Tampere University Hospital approved the study protocol.

Figure 1.  Flow chart of the present study. Eu-tTG = IgA-class tissue transglutaminase antibody (Eu-tTG® umana IgA); Celikey tTG = IgA-class tissue transglutaminase antibody (Celikey® Tissue Transglutaminase IgA Antibody Assay); EMA = immunoglobulin A (IgA)-class endomysial antibody; a = Not including non-participants in the primary study (n=783) or participants with no sera available for the purposes of the current study in 2001 (n=227); b = Cases with previous coeliac disease or dermatitis herpetiformis diagnosis (n=3).

Measurement of antibodies

Serum samples from the participants were stored at −20°C, and altogether 6,987 sera were analysed for immunoglobulin A (IgA)-class tissue transglutaminase antibodies in the first stage of screening (Eu-tTG® umana IgA, Eurospital S.p.A., Trieste, Italy; abbreviated as Eu-tTG) in 2001 (Figure 1). Further, positive sera were analysed parallelly for both IgA endomysial antibodies (abbreviated as EMA) and using another IgA-class tissue transglutaminase antibody kit (Celikey®, Phadia, Freiburg, Germany; abbreviated as Celikey tTG) (Figure 1). The endomysial antibodies were defined by a standardized and validated indirect immunofluorescence method using human umbilical cord as antigen, and a characteristic staining pattern at a serum dilution 1:≥5 was considered positive [27], [28]. Both commercial tissue transglutaminase antibody kits use human recombinant tissue transglutaminase as antigen, and results are given in arbitrary units (AU). The cut-off point for the Eu-tTG was 7.0 AU/mL and for the Celikey tTG 5.0 AU/mL according to manufacturers’ instructions. These IgA-class tissue transglutaminase and EMA tests have been shown to be valid for coeliac disease, with pooled specificities approaching 100% and sensitivities mainly over 90% in adult populations [24], [29]. In accordance with previous studies [5], [30] the definition of unrecognized coeliac disease was based on a two-stage screening algorithm where cases yielding a positive result for both Eu-tTG and either EMA or Celikey tTG were considered to constitute undetected coeliac disease. The main motivation for the two-stage screening algorithm was to decrease the likelihood of false-positive results and thus the dilution of a real effect. Furthermore, to find a close estimate of the real risk we defined unrecognized coeliac disease by two different antibody tests (Celikey tTG and EMA) in the second stage of screening.

In view of the unexpectedly high number of Eu-tTG-positive sera collected 22 years earlier, we wished to check that the influence of storage on Eu-tTG values was not arbitrary. We therefore also tested 128 (1 in 50) randomly selected Eu-tTG-negative sera for EMA and Celikey tTG (Figure 1). As none were positive, we could ascertain that the likelihood of coeliac disease in Eu-tTG-negative individuals was low (Figure 1). We have previously extensively discussed the possible influence of storage of sera on coeliac autoantibodies [5], [30].

Mortality

Date and cause of death were identified by linking the unique personal identification codes with records from the nation-wide database of Statistics Finland [31]. The principal causes of death were coded either according to International Classification of Diseases (ICD)-8, -9, or -10, depending on the time of death. Follow-up commenced the day the blood samples were drawn in 1978–80, and the study subjects were under surveillance until the end of 2005 or death; this yielded a total follow-up of 147,646 person years. The maximum follow-up time was 28 years. The mortality among antibody-positive cases was compared to that of antibody-negative individuals in the same cohort.

Statistical analysis

Adjusted relative risks (RR) and their 95% confidence intervals (95% CI) for mortality comparing antibody-positive with antibody-negative participants were estimated based on a Cox regression model (Cox 1972). Statistical significance of heterogeneity was tested using the likelihood ratio test based on the model and expressed by P-value. The possible confounding effects of age, sex, education, body mass index, alcohol consumption, smoking, hypertension, serum cholesterol, high-density lipoprotein and triglycerides, diabetes, coronary heart disease, stroke, and cancer were assessed using a series of multivariate models. Stratified analyses were conducted to assess mortality risk also according to the level of antibodies (titres 1:<500 and 1:≥500 in EMA, and <6.4 and ≥6.4 AU/mL in Celikey tTG, divided by medians of positive values) and length of follow-up (1–10 years and >10 years). In addition, potential effect-modifying factors such as age, sex, education, body mass index, alcohol consumption, smoking, hypertension, and cholesterol values were entered into the models. The analyses were performed using SAS 9.1 (SAS Institute, Cary, NC, USA).

Results

Altogether 574 out of the 6,987 analysed serum samples proved Eu-tTG-positive (Figure 1). Seventy-four (53 females, mean age 49 years) out of the Eu-tTG-positive had positive EMA, and 204 (125 females, mean age 59 years) positive Celikey tTG. Thus, 1.1% of the 6,987 analysed samples were EMA-positive and, correspondingly, 2.9% Celikey tTG-positive.

As to the personal characteristics of the screened population, subjects positive in either EMA or Celikey tTG had better cholesterol profiles compared to antibody-negative participants (Table I). In addition, Celikey tTG-positive cases were older, consumed more alcohol, and more likely suffered from diabetes than antibody-negative subjects. Otherwise, no statistically significant differences were detected across antibody status. A total of 3,069 (43.9%) participants out of the 6,987 died during the surveillance.

Table I.  Association of personal characteristics with endomysial (EMA) and tissue transglutaminase (Celikey tTG) antibodies: age- and sex-adjusted distributions or mean values with standard deviations (SD) are shown.

	EMA		Celikey tTG
Variable	Negative (n=6913)	Positive n(n=74)	P-value	Negative (n =6783)	Positive (n =204)	P-value
Men ^a,%	46.3	28.0	0.002	46.3	38.7	0.03
Age ^b, years (SD)	51.1 (14.1)	49.2 (11.8)	0.26	50.8 (14.0)	59.1 (14.2)	<0.001
Intermediate or higher education,%	32.4	36.1	0.47	32.3	32.4	0.98
Body mass index, kg/m^2 (SD)	25.9 (4.1)	25.0 (3.9)	0.07	25.9 (4.1)	25.6 (4.4)	0.32
Smokers,%	23.7	19.6	0.38	23.8	19.0	0.10
Physically active,%	15.3	15.1	0.96	15.3	13.7	0.52
Alcohol consumption, g/week (SD)	45.9 (107.8)	33.8 (80.8)	0.31	45.3 (106.1)	60.8 (149.7)	0.03
Hypertensive,%	22.7	20.2	0.60	22.6	26.5	0.18
Total cholesterol, mmol/L (SD)	6.95 (1.37)	6.36 (1.16)	<0.001	6.96 (1.36)	6.37 (1.36)	<0.001
HDL, mmol/L (SD)	1.70 (0.41)	1.50 (0.32)	<0.001	1.70 (0.41)	1.55 (0.38)	<0.001
Triglycerides, mmol/L (SD)	1.54 (1.08)	1.38 (0.49)	0.19	1.53 (1.08)	1.50 (0.73)	0.67
LDL, mmol/L (SD)	4.56 (1.24)	4.24 (1.04)	0.02	4.57 (1.24)	4.15 (1.18)	<0.001
Diabetes,%	5.7	3.3	0.35	5.5	11.1	<0.001
Coronary disease,%	10.9	5.8	0.14	10.9	10.0	0.66
Stroke,%	1.6	1.7	0.95	1.5	3.1	0.08
Cancer,%	2.2	1.4	0.64	2.2	0.1	0.03

^aAge-adjusted.

^bSex-adjusted.

^bHDL = serum high-density lipoprotein; LDL = serum low-density lipoprotein.

No increased age- and sex-adjusted relative risk of overall mortality could be detected among EMA-positive cases, but there was border-line significant modestly elevated risk of overall mortality in Celikey tTG-positive individuals (Table II). The risk levels remained virtually the same after further adjustment for body mass index, smoking, education, alcohol consumption, hypertension, serum cholesterol profile, diabetes, coronary disease, stroke, and cancer. Nor was the risk statistically significantly different in the first 10 years of follow-up (EMA-positive 0.36 (95% CI 0.12–1.11, P=0.03); Celikey tTG-positive 1.17 (95% CI 0.89–1.54, P=0.26)) or thereafter (EMA-positive 0.95 (95% CI 0.61–1.47, P=0.81); Celikey tTG-positive 1.23 (95% CI 0.98–1.56, P=0.09)). Furthermore, high EMA titre had likewise no influence on risk level (0.74, 95% CI 0.42–1.30, P=0.45), but individuals with high Celikey tTG levels had a border-line significantly elevated risk of overall mortality (1.32, 95% CI 1.00–1.72, P=0.12). No statistically significant interactions between any of the potentially effect-modifying factors and antibody status were noted in the prediction of all-cause mortality (data not shown).

Table II.  Age-, sex-, and multivariate-adjusted relative risks of all-cause mortality between persons with positive and negative endomysial (EMA) and tissue transglutaminase (Celikey tTG) antibodies.

		Mortality, relative risks
Antibody	Deaths/persons at risk	Age- and sex-adjusted (95% CI, P-value)	Multivariate-adjusted ^a (95% CI, P-value)	Multivariate-adjusted ^b (95% CI, P-value)
EMA
Positive	23/74	0.78 (0.52–1.18, 0.22)	0.91 (0.59–1.38, 0.64)	0.85 (0.57–1.29, 0.44)
Negative	3046/6913	1.0	1.0	1.0
Celikey tTG
Positive	128/204	1.19 (0.99–1.42, 0.07)	1.18 (0.99–1.42, 0.08)	1.20 (1.00–1.43, 0.06)
Negative	2941/6783	1.0	1.0	1.0

^aAdjusted for sex, age, education, body mass index, smoking, alcohol consumption, hypertension, serum cholesterol, serum high-density lipoprotein, serum triglycerides, diabetes, coronary disease, stroke, and cancer.

^bAdjusted only for sex, age, education, alcohol consumption, smoking, and diabetes. CI = confidence intervals.

Diseases of the circulatory system and malignant neoplasm were leading causes of death among both the antibody-positive and -negative population, comprising 72.0% of all deaths. Antibody-positive cases ran an increased risk of death from lymphoma (Table III). Equally, the risk estimates for stroke and diseases of the respiratory system were increased in both EMA- and Celikey tTG-positive subjects. The risk estimate for dementia as a cause of death was increased only in Celikey tTG-positive cases (Table III).

Table III.  Age- and sex-adjusted relative risks of cause-specific mortality between persons with positive and negative endomysial (EMA) and tissue transglutaminase (Celikey tTG) antibodies.

	EMA				Celikey tTG
Cause of death ^a (ICD code ^b)	Negative (n=6913)	Positive (n=74)	RR ^c (95% CI) ^d	P-value	Negative (n =6783)	Positive (n=204)	RR ^c (95% CI) ^d	P-value
Malignant neoplasm (C00–C97)	655	5	0.73 (0.30–1.77)	0.47	641	19	0.87 (0.55–1.38)	0.56
Digestive organs (C15–C26)	234	1	0.41 (0.06–2.89)	0.29	223	12	1.55 (0.86–2.78)	0.17
Lymphoma (C81–C88)	20	2	9.51 (2.20–41.22)	0.02 ^f	20	2	2.69 (0.62–11.63)	0.24
Diseases of the circulatory system (I00–I99)	1539	12	0.85 (0.48–1.50)	0.57	1482	69	1.23 (0.97–1.57)	0.10
Ischaemic heart diseases (I20–I25)	930	4	0.49 (0.19–1.30)	0.10	904	30	0.90 (0.62–1.30)	0.56
Stroke (I60–I69)	354	4	1.20 (0.45–3.23)	0.72	334	24	1.82 (1.20–2.76)	0.01^e
Diseases of the digestive system (K00–K93)	94	0	-	-	91	3	0.96 (0.30–3.05)^e	0.95
Endocrine diseases (E00–E35)	42	0	-	-	40	2	1.27 (0.31–5.30)^e	0.75
Diseases of the respiratory system (J00–J99)	260	3	1.47 (0.47–4.61)^e	0.53	246	17	1.76 (1.08–2.90)^e	0.04^f
Dementia (F00–F03, G30, R54)	152	1	0.70 (0.10–5.04)	0.71	142	11	2.31 (1.25–4.28)	0.02^f
Accidents, suicide and violence (S00–S99; T00–T98; V01–V99; W00–W99; X00–X99; Y00–Y98)	160	1	0.63 (0.09–4.50)^e	0.62	158	3	0.64 (0.20–2.01)^e	0.41

^aDue to the low number (0–1) of cases with diseases of the nervous system, infectious and parasitic conditions, musculoskeletal system and connective tissue diseases or specific malignancies as cause of death in the antibody-positive group, the results are not shown.

^bRelevant International Classification of Diseases codes (ICD-8, -9, and -10) were used. Corresponding ICD-10 codes are shown.

^cRelative risk estimated by a Cox regression model.

^dThe analysis was repeated after exclusion of cases with the mentioned illness at the beginning of follow-up, and the results remained virtually the same.

^eExclusion of cases with the defined illness at the beginning of follow-up was not possible.

^fP-value < 0.05.

Discussion

No statistically significantly increased risk of all-cause mortality was detected among coeliac antibody-positive unrecognized coeliac disease cases. This is in contrast to findings in earlier studies in apparently symptomatic clinically detected coeliac disease patients, which have shown an increased risk of overall mortality [12–20]. The association between unrecognized coeliac disease and mortality has previously been approached in only two separate studies, with discrepant results [21], [22]. Metzger and associates could show a 2.5-fold increased risk of overall mortality among tissue transglutaminase antibody-positive cases [22]. Their cases were mostly men, whereas in coeliac disease in general females predominate [32]. Comparable to our findings again, Johnston and colleagues [21] found no excess risk, but their cohort consisted of only few EMA- or antireticulin antibody-positive cases. However, it remains unclear whether a border-line statistically significant 19% increased risk of overall mortality among Celikey tTG-positive cases of our study could reach statistical significance in still larger settings. We hypothesize that any possible difference in risk of mortality between endomysial and tissue transglutaminase antibody-positive individuals might be due to border-line positive tissue transglutaminase antibodies outside celiac disease or reflect different subtypes of coeliac disease.

Relating to cause-specific mortality, malignancies at any site were not in general overrepresented in unrecognized coeliac disease. However, albeit based on few cases, the present study indicated that non-Hodgkin's lymphoma as cause of death was overrepresented in undetected coeliac disease. The association is strongly supported by the previous literature, where a connection between diagnosed coeliac disease and non-Hodgkin's lymphoma has repeatedly been reported [15], [16], [18], [20], [33]. As to diseases of the circulatory system, the association with coeliac disease has remained inconclusive [16], [18], [20], [34], [35]. We could show lower cholesterol levels in antibody-positive compared to antibody-negative individuals. Even though the mechanisms for the phenomenon should be evaluated in further studies, we suggest that it might be due to impaired absorption in the intestine. A favourable cardiovascular risk profile has also previously been connected with undetected coeliac disease [3], thus possibly reducing the risk of diseases of the circulatory system. On the other hand, malabsorption of folic acid followed by hyperhomocysteinaemia might increase the risk of these diseases [36–38] and explain the increased risk of stroke in unrecognized coeliac disease cases in the current study. Furthermore, an excess risk of diseases of the respiratory system in general could be demonstrated in our study, as has previously been reported in diagnosed cases [15], [16]. According to the previous literature, specific illnesses such as tuberculosis [39], other lung cavities [40], and sarcoidosis [41] might be in the background.

We are confident regarding the main results of the present study, as we used an adult-representative population-based cohort with a high participation rate (90%) and a substantial number (147,646) of person years, and further tested all sera from the participants by a two-stage screening-algorithm. In addition, with a cohort study design with historical components we could avoid the possible ethical concern associated with a fully prospective design, i.e. following up diagnosed coeliac disease cases without treatment for decades. Nor would a fully prospective study have yielded results in reasonable time. By our population-based screening we were able to avoid confining participants to the most serious cases as is commonly the case in studies with clinically diagnosed coeliac disease. In addition, we compared antibody-positive and -negative individuals of the same cohort and could thus adjust for several potential confounding factors [42]. However, imperfect sensitivity due to IgA deficiency could slightly dilute the real effect [43]. As to the outcome variable, the causes of death register has included all deaths in Finland since 1936, and good validity and coverage of the data in the register have been reported [31], [44]. In brief, details of our study design strengthen the validity of the present results on the prognosis of the unrecognized section of the coeliac population.

Since it might be asked whether good prognosis as regards overall mortality is due to the diagnosis and treatment of unrecognized cases during surveillance, we evaluated risk estimates in the first 10 years of follow-up and thereafter. We found no increased risk of mortality in the first period of follow-up, where the likelihood of coeliac disease diagnosis most probably remained low, or later on. However, individuals with higher levels of tissue transglutaminase antibodies carried a modestly increased risk of mortality, this possibly being explained by more severe disease in these cases [15], [45–47].

The good prognosis found here as regards overall mortality would suggest that a search for asymptomatic coeliac disease by screening programmes is not warranted even though the mortality risks of specific diseases were increased. Other outcome variables such as fractures and quality of life in undetected disease should still be evaluated in future studies to obtain an overall picture of the entity of undetected coeliac disease.

Acknowledgements

The Coeliac Disease Study Group is supported by grants from the Competitive Research Funding of the Pirkanmaa Hospital District, the Emil Aaltonen Foundation, the Foundation for Paediatric Research, the Yrjö Jahnsson Foundation, the Finnish Coeliac Society, and the Academy of Finland, Research Council for Health. The present study was also funded by the Commission of the European Communities in the form of the Research and Technology Development programme ‘Quality of Life and Management of Living Resources’ (QLRT-1999-00037), ‘Evaluation of the Prevalence of Coeliac Disease and its Genetic Components in the European Population’. The study does not necessarily reflect the current views or future policies of the Commission of European Communities. The authors’ work was independent of the funders. Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.