Are We Close to Predicting the Onset of Celiac Disease in Children?

Predicting Onset Celiac DiseaseCeliac disease (CD) has a prevalence of nearly 1% of many countries’ populations, and its incidence appears to be increasing. The risk of CD is elevated among people with a family history,3 and therefore, testing in this population, even among asymptomatic individuals, has been advocated.

However, the frequency with which to test has not been established, and given the possibility of seroconversion and loss of tolerance to gluten during childhood and beyond, repeat testing may be advisable.

But at what age should testing occur, and how often should it be performed? It is within this context of uncertainty that the study by Meijer et al, in this issue of Gastroenterology, sheds important light.

  • Benjamin Lebwohl, Luigi Greco, 1

In this study, the investigators analyze long-term follow-up data from the PreventCD trial, a randomized trial of infants with a first-degree relative with CD that was designed to test the strategy of low-dose gluten introduction at age 4 months.

The trial did not show that this strategy reduced the risk of CD development, but the abundant data collected during this trial have allowed these investigators to study risk factors for the development of CD among the trial participants.

The present analysis takes into account the child’s

  • age,
  • sex,
  • HLA type,
  • and number of affected relatives,

and, based on these risk factors, provides a prediction score that estimates the cumulative risk of CD in the ensuing years.

Strengths of the study include its prospective design, comprehensive data collection, and the use of an independent validation cohort outside of the PreventCD trial to test their prediction score.

Beyond merely providing general risk stratification, the investigators have translated this into clinical applicability: They have developed an easy-to-use web-based calculator (available at that provides useful data for families and health care providers, including absolute risks and recommendations regarding testing frequency based on the individual patient’s risk profile.

In addition to providing a recommended timetable for screening, these study results also expand the way we think about genetic test results in CD.

HLA testing in this setting has historically been performed primarily due to its excellent negative predictive value. Because HLA DQ2 and DQ8 are present in nearly 100% of people with CD, the primary value of its use has been in ruling out CD when an individual is found to have neither haplotype.

As such, HLA testing has been useful in certain scenarios, such as

  • patients with discrepant serologic and histologic results,
  • those who already started a gluten-free diet before clinical assessment, or, as is the case in the present study,
  • children with a family history of CD.

However, in this last category, it appears that we should be paying attention to details of HLA results beyond its negative predictive value.

The results of well-characterized studies, including those using The Environmental Determinants of Diabetes in the Young (TEDDY) cohort, show the specific HLA haplotype matters, and there is a gradient of risk, with DQ2.5 homozygosity at the top of this gradient.

As a result of the risk calculator developed in the present study, we can now use the details of our patients’ HLA haplotype to develop a rational testing strategy and think beyond the dichotomy of “positive” or “negative” genetic testing in CD.

This study and its risk calculator have several limitations.

  • The analysis did not include data beyond age 10,
  • and the calculator only provides incidence data and screening recommendations through age 8.
  • Relatively sparse testing of cohort participants in the latter years may also underestimate CD risk.

Beyond these issues, a larger question is whether asymptomatic children should be screened.

This debate has not been resolved. Guidelines from the North American Society for Pediatric Gastroenterology, Hepatology, and Nutrition (NASPGHAN) recommend that asymptomatic children in high-risk groups (including first-degree relatives) be screened,

– but –

the United States Preventive Services Task Force concluded that the evidence is insufficient to warrant recommending for or against screening asymptomatic individuals, including children.

But among those families and practitioners who believe that the benefits of screening justify its practice, the risk calculator provides a path and schedule.

The risk calculator takes into account sex, HLA risk group, and the number of affected relatives. But even those in the highest risk categories do not have a risk of CD that is >50%.

What are the other factors that influence CD risk?

Of the environmental risk factors that have been proposed, the quantity of gluten consumption during early childhood is perhaps the most interrogated suspect.

Although the PreventCD trial did not show an effect of low-dose gluten on reducing the risk of CD, the present analysis found an association between higher gluten consumption in early childhood and an increased risk of subsequent CD; similar findings have been reported in multiple observational studies in recent years.

Despite the accumulating data that quantity of gluten consumption can influence CD, there is also evidence that risk of CD is influenced, perhaps even determined, before gluten is introduced.

One study showed infants with a DQ2 haplotype had distinct a gut microbial community at age 1 month, associated with CD risk, compared with infants without a CD-compatible haplotype.

Similarly, infants who subsequently developed CD were found to have increased serum cytokines at age 4 months, before gluten introduction.These findings raise the possibility that efforts to prevent CD may require more than modifying gluten exposure.

While such efforts are ongoing, the study by Meijer et al may allow us to more effectively monitor for the development of CD in young children with a family history.

This is particularly important given rising rates of CD today. In this trial, CD developed in 135 of 944 participants (14%), and given the rising incidence and prevalence of CD, with longer follow-up, it is anticipated this number will be higher.

The prevalence of CD among children in Colorado has been estimated to be as high as 3% by age 15 years; monitoring tools cannot come soon enough.