Healthcare organizations adopt performance improvements at different rates because of differences in incentives, organizational processes, and management styles [1]. An important complicating factor in complex healthcare systems like the United States (US) is the lack of full transparency in costs, performance metrics, and clinical outcomes. As such, better or different clinical strategies may not be implemented widely, rapidly, or at all [2].
The adoption of novel practices or performance improvements may face a comparable set of resistance forces at the national or international levels. These forces accelerate the adoption of processes deemed favorable to individual providers or healthcare systems (e.g., safer medications or higher reimbursements), and decelerate the adoption of disruptive processes deemed unfavorable (e.g., elimination of revenue-generating procedures or adoption of standardized protocols). One notable example of delayed clinical implementation at the international level is the diagnostic management of pediatric Celiac Disease (CD).
The policies and positions of the European (ESPGHAN) and North American (NASPGHAN) Societies for Pediatric Gastroenterology, Hepatology and Nutrition shape the practice of pediatric gastroenterology. These sister societies frequently issue consensus guidelines, and until 2012 had equivalent diagnostic guidelines for the management of CD. In 2012, however, ESPGHAN issued a set of revised guidelines that allowed a “no-biopsy” diagnostic pathway patients with a serum immunoglobulin A anti-tissue transglutaminase antibody (tTG) titer greater than 10 times the upper limit of normal (>10x ULN) [3]. Support for this revised position included a detailed analysis of the clinical evidence [4], the opinion of practicing physicians [5], and the results from preliminary clinical testing in a variety of conditions [6, 7].
At the time of the publication of ESPGHAN criteria, North American experts appropriately suggested that “there is still a long way to go but we are headed in the right direction” towards no-biopsy diagnosis of CD in any patient [8]. Despite multiple opportunities to reach consensus since 2012, NASPGHAN and the American College of Gastroenterology continue to maintain biopsy as a required part of the diagnosis for every suspected case of CD [9, 10]. The American Gastroenterological Association clinical update recently discussed both European and North American approach [11], but did not adopt a specific position regarding the no-biopsy approach in any patient group. Since 2012, the European experts reaffirmed and extended their position that pediatric CD can be diagnosed without biopsy in a selected group of children by following the recommended guidelines [12].
Given the substantial costs and health implications of endoscopy with biopsy in children, we explore the value of the information provided by biopsy in children with high titer serum tTG results in a large North American referral center. We show that with high pre-test probability of CD based on serum tTG values, duodenal biopsy provides no additional diagnostic value for CD, consistent with ESPGHAN findings. Moreover, biopsy counter-intuitively introduces diagnostic uncertainty in a number of patients necessitating further clinical action or follow-up. We briefly explore the economic consequences of biopsies and present a system dynamics framework to understand feedback mechanisms that enforce the status quo in North America. The remainder of this background provides relevant information about the pathophysiology and diagnosis of CD, including the evolution of diagnostic recommendations from by ESPGHAN and the North American response.
Pathophysiology of CD
CD has a prevalence of 0.4-1% [13] and is in the differential diagnosis of children with any gastrointestinal symptom, particularly with predisposing conditions, including autoimmune disease, diabetes, Down syndrome, and family history [14]. Serological screening is the first line of action for evaluation of any patient with clinical suspicion of CD [3, 9, 10, 13, 15-18]. Patients with positive serology are typically referred for upper gastrointestinal (UGI) endoscopy and biopsies. Since CD is a small intestinal disease, duodenal histological abnormalities are considered the hallmark of active disease [7, 16, 17, 19-25]. Small intestinal abnormalities in CD were described in 1960s [26-31], and widespread availability of endoscopy made duodenal biopsy the de facto diagnostic standard. For decades, histology served as the only reliable biomarker for the disease, became known as the “gold standard,” and has remained such in spite of significant advances in laboratory testing and endoscopic imaging.
In spite of its central role in diagnosis, biopsy has well-known limitations [24, 32]. Overlap exists between histopathological findings in CD and other conditions ranging from infections to systemic disorders [24]. Writing on behalf of Gastrointestinal Pathology Society and the Association for Study of Celiac Disease, Robert et al. (2018) concluded that "correlation of histologic findings in duodenal biopsies with patient demographics, symptoms, medication use, evidence of H. pylori infection, and laboratory data, especially serological and genetic tests for Celiac Disease is required for correct diagnosis." Thus, consideration of histopathology as the gold standard is not supported in practice by the need for extensive clinical correlation to reach a correct diagnosis. The widely-used Marsh histological classification acknowledges the presence of a histological spectrum, emphasizing less than perfect sensitivity and specificity of biopsy [24, 33, 34]. Importantly, all classical descriptions of CD histopathology relied on gluten-sensitivity as the definitive evidence of CD, rather than proposing the presence of pathognomonic histological features [30, 33, 35]. Pathologically, CD may show: (i) no specific histopathological findings, (ii) classical histopathology of active CD, or (iii) concurrent or superimposed confounding pathologies. Although duodenal biopsy can provide confirmation of CD if and when classical features are present, the overall performance characteristics of biopsy remain poorly quantified and variable because of histological overlap between multiple different inflammatory entities [reviewed in 24].
A key issue limiting the reliability of biopsy is histological variability in tissue expression of CD [36-40]. This biological variability that can result in diagnostic uncertainty is further confounded by well-known tissue processing and interpretive errors in pathology, and biopsies in 4%-30% of patients may be inadequate due to technical issues or interpretive disagreements [41-47]. Thus, recognizing that negative or non-diagnostic duodenal biopsies do not exclude CD [9, 15], practice guidelines suggest that follow-up endoscopy with additional biopsies may be justified or necessary in some patients with clinical and serological evidence of CD (i.e., high pre-test probability) for whom the laboratory reports a negative initial biopsy result [9, 14, 15, 17]. Longitudinal studies have also demonstrated histological evolution over time in patients who carry the diagnosis of CD based on clinical, serological and genetic data [48]. In these patients, duodenal histology at presentation can be non-diagnostic, suggesting that biopsy is an inherently suboptimal test in early CD.
European Movement Towards No-Biopsy
Acknowledging that abnormal histology is a biomarker for CD, one can appreciate the potential existence of other biomarkers (e.g., imaging, serologies or genotypes) with performance characteristics similar to, or possibly better than biopsy. Unlike histopathology, some biomarkers (e.g, genotypes) are independent of age and exposure to gluten, and therefore more generally applicable as a diagnostic tool.
An equally important concept is the probabilistic nature of all diagnostic information [49]. For example, diabetes confers 5-10% probability of CD [50], and a first-degree relative with CD is associated with 7.5% probability of CD [51]. Together, these prior probabilities imply that a patient with diabetes and an affected first-degree relative has a 7.5%-16% probability of CD, depending on the level of linkage between these risk factors. Similar arguments can be made for Down syndrome, associated with CD in up to 18.6% [52], and for multiple other conditions highly correlated with CD [14, 32]. In these circumstances when the pre-test probability of CD is high, if serum tTG level rises from normal on gluten-free diet to >10x ULN after exposure to gluten, there is virtually no alternative diagnosis other than CD, regardless of any biopsy findings. The immediate utility of this probabilistic approach has been shown by others [53]. Therefore, the key policy issue is defining the population(s) in which additional testing (e.g., biopsy) provide diagnostic value and for which the benefits from the information exceed the costs of obtaining it [54].
Based on the Bayesian concept of essentially 100% positive predictive value for CD in a (i) symptomatic child, with (ii) serum tTG >10x ULN, and (iii) positive results of a second Celiac-specific test, ESPGHAN concluded that CD may be diagnosed without biopsy provided that (iv) signs and symptoms subside on gluten-free diet (i.e., establishment of gluten-sensitivity) [3]. These guidelines reaffirmed clinical experience suggesting that biopsy is not always necessary in patients with high pre-test probability of CD [20, 55]. The guidelines further recognize that histological variability can lead, and has led, to the need to perform multiple biopsies (with the additional procedure costs and risks) in individual patients with high-probability of CD who have indefinite or otherwise non-diagnostic biopsies at presentation [36-40].
Since 2012, the ESPGHAN no-biopsy approach has been evaluated in a variety of settings, demonstrating the overall effectiveness of the strategy [6, 22, 25, 41, 56-59]. These studies have shown opportunities for improvement, but none presented a significant challenge to the core concept that a sub-population of patients exists in which CD can correctly and confidently be diagnosed without biopsy. In one such study, the no-biopsy algorithm showed a positive predictive value of 0.988 and a negative predictive value of 0.958 [41]. This and similar recent observations [60, 61] led to reaffirmation and further extension of no-biopsy approach to include asymptomatic children as well [12].
North American Response
In spite of years of accumulated evidence, debate continues in the US about the adoption of any no-biopsy approach [10, 11, 14, 16-18, 23]. Published practice guidelines require a positive concordance between serologies and biopsy for the diagnosis of CD, and recommend obtaining multiple biopsies from distal duodenum and the duodenal bulb regardless of the pre-test probability of the disease [9, 10, 15]. A recent clinical practice guideline discussed a “biopsy-avoiding” approach and acknowledged the existence of patients in which the pre-biopsy probability of CD is “virtually 100%,” but did not specifically endorse a no-biopsy protocol [11]. Confirming the validity of the ESPGHAN guidelines in other populations has been identified as a critical need because of potential clinical differences between different patient populations [8].
An important concern raised by the proponents of an all-biopsy approach (i.e., biopsy every suspected CD case) is the uncertainty about tTG assay performance [8, 9, 14, 23, 62]. These include differences in platforms, technologies, and lack of harmonization among different laboratories that prevent cross-institutional comparison of laboratory results. Others point out a missed opportunity to diagnose incidental disorders as a disadvantage of the no-biopsy approach [8, 9, 23] without providing any formal policy, cost-benefit, or value-of-information analysis as support. Some clinicians express concern that a gluten-free diet may be cumbersome, expensive, and adversely impact the quality of life of the individual. They require confirmation of the diagnosis at the highest level of certainty before recommending a lifelong treatment [9, 62]. Thus, they implicitly value the benefits of biopsy more than its costs.