The concept of a ‘celiac iceberg’ has been important in drawing attention to a large, unrecognized group of patients with CD who do report symptoms considered ‘typical’ of CD . Investigators have proposed expansion of the ‘iceberg’ to encompass patients who are genetically susceptible to CD, but show only raised IEL counts or an isolated abnormal CD-specific serology and normal intestinal histology [30–32]. Consequently, there is considerable uncertainty regarding the true extent of gluten-mediated disease in the community. The recent revision of diagnostic criteria proposed by the ESPGHAN Working Group seeks to address these developments, and increase the contribution of serology and HLA-DQ genetic tests to diagnosis [33–36].
However, Walker et al. highlighted the difficulties in defining the community prevalence of CD . From 1,000 randomly selected Swedish adults who underwent diagnostic gastroscopy, 33 were TG2 IgA-positive, and 16 of these were also EMA-positive. These 16 subjects who were seropositive for both TG2 IgA and EMA were considered to have CD, but only 12 of the 16 EMA subjects showed both villous atrophy and increased density of IELs; of the other 4 patients, 3 showed only increased IELs and the fourth had normal histology. In addition, two of the TG2 IgA+ EMA− subjects also showed partial villous atrophy and raised IEL numbers. Walker et al. speculated that if increased TG2 IgA with normal intestinal histology or raised IEL numbers without increased TG2-IgA reflected gluten-mediated pathology, then the community prevalence of CD might be considerably greater than their estimate of almost 2% .
Consistent with a previous study from Australia and several from Europe and North America [14, 20, 37, 38], over 98% of locally recruited volunteers with ‘biopsy-confirmed’ CD possessed HLA-DQ2.5, DQ8, or DQ2.2. In addition, we found that absence of HLA-DQ2.5, DQ8, or DQ2.2 in ‘biopsy-confirmed’ cases predicted incorrect diagnosis. Building upon this observation, we developed a novel method to estimate the prevalence of CD in two large age-stratified, randomly selected community cohorts, based on the relative enrichment for HLA-DQ alleles conveying genetic susceptibility to CD in the group of individuals with abnormal CD-specific serology tests. If, as some previous studies have done, we had assumed that all TG2 IgA-seropositive individuals with HLA-DQ2.5, DQ8, or DQ2.2 did have CD, then we would have estimated community prevalence to be 3.1% in women and 4.8% in men. The novel serogenetic approach we developed potentially overcomes not only the need for biopsy but also performance variation between different TG2 IgA ELISAs when estimating the prevalence of CD in the community. In this study, we estimated that 1.3% of men and women in the community had unrecognized CD using TG2 IgA serology. However, our finding that two TG2 IgA-seronegative women who were positive in the composite TG2/DGP IgA/IgG screen were subsequently confirmed to have CD on small bowel histology added support to the higher prevalence figure of 1.9% for women, estimated using the composite TG2/DGP IgA/IgG screen.
Clinical follow-up confirmed 10 new cases of unrecognized CD in addition to 7 cases diagnosed through standard medical care before or during the current prospective study. As anticipated, the biopsy-confirmed prevalence of CD of 0.7% in women and 0.6% in men was substantially less than by the serogenetic method. However, for a variety of reasons, seven additional cases with persistent serological abnormalities did not undergo gastroscopy, and more than half of the cases who were positive in the composite TG2/DGP IgA/IgG screen were not available for follow-up or had no additional investigation. Medical review of seropositive subjects by their primary care physicians also highlighted the persisting misconception that CD only presents with classical symptoms. Although use of a gluten-free diet has become much more popular since the enrolment sera were collected, none of the subjects with raised CD-specific serology reported having adopted gluten-free diet, apart from those actually diagnosed with CD. However, it is possible that deliberate or inadvertent reduction of dietary gluten by some formerly seropositive subjects could have resulted in normalization of serological abnormalities.
Testing for CD is increasing. Claims for serological testing for CD reimbursed by Medicare Australia in the State of Victoria rose from 9.3 per 1,000 population in 2004 to 14.4 per 1,000 population in 2009; the total number of claims during this 6-year period was 72.4 per 1,000 population. Females and males aged over 15 years accounted for 63% and 23%, respectively, of all Medicare claims for coeliac disease serology tests in 2010. Between 2004 and 2009, 4 of the estimated 12 to 26 unrecognized cases of CD in our female cohort and 2 of the estimated 12 to 16 unrecognized cases in our male cohort were diagnosed through standard medical care. Over time, as testing increases, the total number of undiagnosed adult cases of CD would be expected to fall steadily, particularly in women. The combined effects of increased testing and increased diagnoses would eventually reduce the pre-test probability of CD when testing patients in primary care. TG2 IgA serology is widely recommended as the initial investigation when CD is considered, but based on serogenetic data from the current study, fewer than one in three women and one in five men with abnormal TG2 IgA in the community would be expected to have unrecognized CD. Hopper et al. also reported similarly high rates of false positives using a different TG2 IgA assay in a group of 2,000 patients referred for gastroscopy . If gastroscopy with intestinal biopsy continues to be recommended as the next step after positive TG2 IgA serology, many unnecessary gastroscopies will be performed, at considerable expense and inconvenience for patients.
To address this problem, we modeled the practical and financial effects of including confirmatory serology and genetic testing after initial testing and before ‘definitive’ small bowel histology. Inclusion of genetic testing after either TG2 IgA or composite TG2/DGP IgA/IgG serology reduced gastroscopies and overall costs incurred according to the current Medicare Australia reimbursement schedule. Inclusion of confirmatory serology after initial testing with composite TG2/DGP IgA/IgG serology had a similar effect by reducing gastroscopies without appreciably affecting detection rates, and also lowering overall costs according to pricings in Australia. The applicability and financial effect of the diagnostic algorithms are likely to vary between countries and healthcare systems, depending on local laboratory facilities and costing. However, the community rates of ‘positive’ and ‘negative’ serologies and the genetic susceptibility data presented in the current study can be combined with relevant local costs for laboratory assays and procedures to predict which diagnostic pathway might be the most relevant and cost-effective in a particular setting.
Although it would be premature to formally propose a ‘single’ diagnostic flow chart, the present study highlights the effects of separating initial laboratory testing by serology from a second set of laboratory tests in order to further increase the likelihood of CD before proceeding to endoscopic biopsy and histology, which is definitive but also the most expensive and intrusive investigation used in diagnostic investigations for CD. Shifting costs from gastroscopy to laboratory testing promises to reduce overall costs and may be more acceptable to patients, particularly if gastroscopy is difficult to access or even undesirable.