Associations Between Celiac Disease and Type 1 Diabetes
Associations Between Celiac Disease and Type 1 Diabetes
We searched PubMed for CD ('coeliac' or 'celiac') AND 'diabetes mellitus' published between 1 January 2000 and 1 May 2014. We decided to restrict our search to this time period since only around year 2000 did research on screening with tissue transglutaminase antibodies (TTG) come into use following the discovery of tissue transglutaminase as the autoantigen of CD.
We restricted our search to English-language publications. The search was conducted by JFL with support from PE. Titles and abstracts were read for 587 papers. Through Web of science (http://thomsonreuters.com/thomson-reuters-web-of-science/) we also examined papers that had cited the studies by Collin et al., Hansen et al. and our own paper on CD and subsequent T1D.
Prevalence of CD in Patients With T1D. We restricted our review to papers with ≥100 T1D individuals undergoing screening for CD. All potential studies were screened by JFL. PE verified key data. Disagreements were resolved by consensus. Forty papers (including one letter to the editor) were read in detail and 27 papers qualified for our meta-analysis (Table 1). No author was contacted, and where CIs for the proportion of individuals with CD was missing, we instead used WolframAlpha (http://www.wolframalpha.com) to calculate upper and lower 95% CIs and standard errors based on the proportions given in each paper.
We identified no duplicate publications (author names, similar sample size, and country were used to rule out duplication).
Relative Risk of T1D in Patients With a Diagnosis of CD. We only identified one study where relative risks had been calculated. Hence, no meta-analysis was performed for this outcome.
We retrieved information on (i) year of publication, (ii) age group (children, adults, both children and adults), (iii) country, (iv) proportion of antibody-positive individuals that underwent small intestinal biopsy. If only a small proportion of individuals who are positive for coeliac antibodies [endomysial antibodies (EMA) or TTG] undergo biopsy, the prevalence of biopsy-verified CD will naturally be lower than in a setting where everyone with positive coeliac serology undergoes a biopsy.
Although PubMed classifies children as individuals aged 0–18 years (and adults 19+ years), other definitions of age-groups occur and we cannot rule out that also individuals aged 19–21 years have been included as children in some studies (definitions were not always clear). We did not grade studies according to diagnostic quality but certain aspects are discussed in the discussion.
We performed our meta-analysis by computing the prevalence of CD in T1D diabetes using a random-effects model. 95% CIs were calculated.
Heterogeneity was calculated using I squared (I).
In order to investigate heterogeneity, we also examined prevalences by subgroups of study size, continent (geography: Europe, North America, Other), age and proportion of serologically positive individuals undergoing small intestinal biopsy (≤99% vs. 100%), proportion of females (<50% vs. ≥50%), and whether all patients during a time frame (i.e. consecutive patients) were tested or not. To examine the linear relationship between CD prevalence and (i) study size, (ii) year of publication, (iii) proportion of serologically positive T1D patients undergoing biopsy, (iv) CD prevalence in the general population, and (v) Gross Domestic Product (GDP) per capita, we also carried out meta-regressions. Due to lack of data on country-specific biopsy-verified prevalence of CD we used proxies (data from the Netherlands for Belgium; and from the US for Canada). For Italy and the UK, we defined CD according to the 'prevalence B criteria' in the multi-European screening study by Mustalahti et al. (and where studies consisted of both children and adults we used a mean prevalence: Italy: 0.9%), while for Sweden we based our general population-based estimate on the Kalixanda study (1.8%).
Considering the large size of the Franzese et al. paper (n = 8717 patients), a sensitivity analysis was carried out without this study.
When authors have reported that the histopathology was graded according to the revised ESPGHAN criteria (and the report was published before 2012), we have assumed that villous atrophy (Marsh III) was required for the diagnosis of CD. However, this can be interpreted differently, as for instance Bhadada et al. claimed to use the modified ESPGHAN criteria and then nevertheless accepted Marsh 1 as proof of CD. This will naturally increase the prevalence of CD in T1D. In a sensitivity analysis, we however, restricted our study to studies with Marsh stage III.
When there were no data on whether seropositive individuals underwent biopsy we assumed that all individuals had a biopsy (e.g.)
As transiently positive coeliac serology may occur in T1D patients, we used biopsy-verified CD as our outcome. While biopsy is still recommended for the diagnosis of CD in adults, the European organisation for paediatric gastroenterology (ESPGHAN) allows for biopsy-free diagnosis of CD since 2012 in symptomatic children >2 years, with DQ2/DQ8+, TTG >10 times normal and EMA+. However, in the ESPGHAN recommendations for children with T1D, small intestinal biopsy is still recommended. Traditionally, CD has been equal to histopathology Marsh grade III, but in recent years, also milder histopathology has been deemed consistent with CD in patients with positive CD serology. In the current meta-analysis, we did not differ between Marsh III and milder histopathology for the diagnosis of CD.
We used Stata 13 for all analyses.
Methods
Search
We searched PubMed for CD ('coeliac' or 'celiac') AND 'diabetes mellitus' published between 1 January 2000 and 1 May 2014. We decided to restrict our search to this time period since only around year 2000 did research on screening with tissue transglutaminase antibodies (TTG) come into use following the discovery of tissue transglutaminase as the autoantigen of CD.
We restricted our search to English-language publications. The search was conducted by JFL with support from PE. Titles and abstracts were read for 587 papers. Through Web of science (http://thomsonreuters.com/thomson-reuters-web-of-science/) we also examined papers that had cited the studies by Collin et al., Hansen et al. and our own paper on CD and subsequent T1D.
Study Selection
Prevalence of CD in Patients With T1D. We restricted our review to papers with ≥100 T1D individuals undergoing screening for CD. All potential studies were screened by JFL. PE verified key data. Disagreements were resolved by consensus. Forty papers (including one letter to the editor) were read in detail and 27 papers qualified for our meta-analysis (Table 1). No author was contacted, and where CIs for the proportion of individuals with CD was missing, we instead used WolframAlpha (http://www.wolframalpha.com) to calculate upper and lower 95% CIs and standard errors based on the proportions given in each paper.
We identified no duplicate publications (author names, similar sample size, and country were used to rule out duplication).
Relative Risk of T1D in Patients With a Diagnosis of CD. We only identified one study where relative risks had been calculated. Hence, no meta-analysis was performed for this outcome.
Data Items and Risk of Bias
We retrieved information on (i) year of publication, (ii) age group (children, adults, both children and adults), (iii) country, (iv) proportion of antibody-positive individuals that underwent small intestinal biopsy. If only a small proportion of individuals who are positive for coeliac antibodies [endomysial antibodies (EMA) or TTG] undergo biopsy, the prevalence of biopsy-verified CD will naturally be lower than in a setting where everyone with positive coeliac serology undergoes a biopsy.
Although PubMed classifies children as individuals aged 0–18 years (and adults 19+ years), other definitions of age-groups occur and we cannot rule out that also individuals aged 19–21 years have been included as children in some studies (definitions were not always clear). We did not grade studies according to diagnostic quality but certain aspects are discussed in the discussion.
Summary Measures, Analysis Method and Heterogeneity
We performed our meta-analysis by computing the prevalence of CD in T1D diabetes using a random-effects model. 95% CIs were calculated.
Heterogeneity was calculated using I squared (I).
In order to investigate heterogeneity, we also examined prevalences by subgroups of study size, continent (geography: Europe, North America, Other), age and proportion of serologically positive individuals undergoing small intestinal biopsy (≤99% vs. 100%), proportion of females (<50% vs. ≥50%), and whether all patients during a time frame (i.e. consecutive patients) were tested or not. To examine the linear relationship between CD prevalence and (i) study size, (ii) year of publication, (iii) proportion of serologically positive T1D patients undergoing biopsy, (iv) CD prevalence in the general population, and (v) Gross Domestic Product (GDP) per capita, we also carried out meta-regressions. Due to lack of data on country-specific biopsy-verified prevalence of CD we used proxies (data from the Netherlands for Belgium; and from the US for Canada). For Italy and the UK, we defined CD according to the 'prevalence B criteria' in the multi-European screening study by Mustalahti et al. (and where studies consisted of both children and adults we used a mean prevalence: Italy: 0.9%), while for Sweden we based our general population-based estimate on the Kalixanda study (1.8%).
Considering the large size of the Franzese et al. paper (n = 8717 patients), a sensitivity analysis was carried out without this study.
When authors have reported that the histopathology was graded according to the revised ESPGHAN criteria (and the report was published before 2012), we have assumed that villous atrophy (Marsh III) was required for the diagnosis of CD. However, this can be interpreted differently, as for instance Bhadada et al. claimed to use the modified ESPGHAN criteria and then nevertheless accepted Marsh 1 as proof of CD. This will naturally increase the prevalence of CD in T1D. In a sensitivity analysis, we however, restricted our study to studies with Marsh stage III.
When there were no data on whether seropositive individuals underwent biopsy we assumed that all individuals had a biopsy (e.g.)
As transiently positive coeliac serology may occur in T1D patients, we used biopsy-verified CD as our outcome. While biopsy is still recommended for the diagnosis of CD in adults, the European organisation for paediatric gastroenterology (ESPGHAN) allows for biopsy-free diagnosis of CD since 2012 in symptomatic children >2 years, with DQ2/DQ8+, TTG >10 times normal and EMA+. However, in the ESPGHAN recommendations for children with T1D, small intestinal biopsy is still recommended. Traditionally, CD has been equal to histopathology Marsh grade III, but in recent years, also milder histopathology has been deemed consistent with CD in patients with positive CD serology. In the current meta-analysis, we did not differ between Marsh III and milder histopathology for the diagnosis of CD.
Statistics Software
We used Stata 13 for all analyses.
Source...