EDITORIAL Eosinophilic Esophagitis—Emerging Epidemic or Misdiagnosed Malady? osinophilic esophagitis (EoE) is a recently recognized, clinicopathologic disorder characterized by esophageal mucosal eosinophilia in association with dysphagia. EoE was first described in a few case reports in adults in the 1980s. It was first characterized as an entity seemingly distinct from gastroesophageal reflux disease in 2 small case series of adults in the early 1990s. Ironically, over the next decade, EoE was mainly viewed as a childhood disease, owing to large cohorts and prospective trials emerging from pediatric centers in the United States. This earlier appreciation by pediatricians has been partly attributed to their more routine clinical practice of obtaining esophageal biopsies at the time of upper endoscopic procedures.1 A more recent single-center study reported that 2.3% of adult patients presenting for endoscopy have esophageal eosinophilia and of these w79% had definitive or probable EoE.2 Recent pediatric and adult prevalence estimates of 43–91 cases per 100,000 have been reported from referral centers caring for patients within restricted catchment areas, including Hamilton county, Ohio, Olmstead county, Minnesota, and Olten county, Switzerland.3–5 The highest prevalence estimate reported of 1 case per 1000 is notable in that the cohort, unlike other population studies, was limited to pediatric cases.5 The article by Dellon and coauthors6 sheds new light on the prevalence of EoE in the United States following the introduction of a dedicated International Classification of Diseases–Ninth Revision (ICD-9) code in 2008 (530.13). It should be noted that the approval of the ICD-9 code was greatly facilitated by patient advocacy (American Partnership for Eosinophilic Disorders). As the authors point out, conducting robust epidemiologic studies in the United States is often difficult because of the lack of a centralized medical record and fragmented health care. Previous work by the same group recently validated the instance of the specific ICD-9 code as specific for identifying cases of EoE (specificity >99% and sensitivity of 37%).7 The current paper applied this definition to a large health insurance claims database that has been previously shown to be representative of the US commercially insured population. Using the new ICD-9 code, the US prevalence estimate for EoE is 56.7/100,000 persons (w152,152 cases). The authors should be complimented on the robust study design, which attempted to minimize misclassification bias by limiting their prevalence estimates to only those patients that were continuously enrolled in the insurance database. In addition, they conducted adequate sensitivity analyses with more restrictive coding definitions of EoE, with the most conservative approach having minimal effect on the prevalence estimates. So, what are we to make of this new US prevalence estimate for EoE? The estimate is on the same order of
E
magnitude as previously published, single-center population studies.2–5 The estimate is likely more precise than prior studies based on the significantly larger sample size. Moreover, the insurance database was equally representative across regions of the country and age groups inclusive of both adults and children. In fact, one of the more notable findings reported is that in the time period studied, 75% of cases were adults. This reflects a paradigm shift in our thinking of EoE; EoE has returned to its origins as a predominantly adult disease. Although more precise, challenges certainly exist regarding the accuracy of a new ICD-9 code to define prevalent cases of a new disease. Esophageal eosinophilia is a histologic finding that also occurs in gastroesophageal reflux disease and proton-pump inhibitor–responsive eosinophilia. Other disease states such as eosinophilic gastroenteritis and connective tissue disorders, albeit rare, can be associated with esophageal eosinophilia. The actual diagnosis of EoE relies both on histopathology and clinical context as the disease definition continues to evolve.8,9 More specific biomarkers of the disease are being studied that may refine the diagnosis. Dellon and coauthors suggest that the ICD-9 code was specific for EoE; however, their validation was performed at a large tertiary referral center that is well known for treating esophageal disorders. Whether or not other practice settings are accurately classifying patients with EoE (and hence using the ICD-9 code correctly) is not known. Greater awareness of the existence of the ICD-9 code in an EoE tertiary center may overestimate the sensitivity whereas greater utilization of up-to-date diagnostic criteria in a tertiary center may overestimate the specificity of the code. In addition, challenges of using ICD-9 codes to define disease states has been documented in numerous instances of diseases with much longer histories than EoE including diabetes and gastroesophageal reflux disease.10–12 Patients may be assigned suspected diagnoses (ICD-9 codes) based on symptom presentation or endoscopic findings prior to actual confirmation of the disease. Another important limitation to the generalizability of the prevalence estimates in the current study pertains to the insurance database. The prevalence of EoE in the insured population may not accurately reflect the demographic differences in the uninsured. Furthermore, the database excluded patients older than age 65 and therefore cannot be extrapolated to the entire US population. Because of the relatively short existence of the EoE ICD code, there is significant opportunity among the gastroenterology, allergy, and pathology communities to ensure that it is used as accurately as possible. As the world of “big data” and use of electronic health records expands for research and clinical practice, data-mining techniques will allow for large population-based studies comparing treatments and monitoring patient outcomes across practice settings and regions. Incident user study designs are useful in limiting Clinical Gastroenterology and Hepatology 2014;12:597–598
598
Gawron and Hirano
bias in retrospective studies aimed at assessing comparative effectiveness of different therapies. As the therapeutic armamentarium for EoE increases and becomes more variable, it is important that researchers be able to accurately define cases and controls in retrospective epidemiologic and therapeutic comparison studies. The high reported specificity for the ICD-9 needs validation in cohorts outside of tertiary networks. Furthermore, the low sensitivity (37%) the authors quote from prior work validating the use of the ICD-9 code with their institutional database of confirmed EoE cases7 implies there are currently many false negatives. Thus, any future epidemiologic studies using this code should be very cautious in defining control groups as decreased sensitivity will not substantially bias relative risk as long as misclassification is nondifferential.13 The authors argue that because of the low sensitivity (ie, w37%) of using the ICD-9 code to identify EoE, the true prevalence of EoE could be as high 411,222 cases of EoE in the United States. EoE is currently considered a rare or “orphan” disease as defined by the Rare Disease Act of 2002 (disease affecting less than 200,000 individuals).14 This affords research opportunities via the Office of Rare Diseases Research and affiliated Rare Disease Clinical Research Network. Thus, accurate prevalence estimates have very important consequences for EoE patients, pharmaceutical companies, and researchers. At this time, we would caution against extrapolating population prevalence for EoE until longer-term data and further validation is available. In summary, Dellon and colleagues have provided a more precise and probably more accurate US population prevalence estimate of EoE. A disease that was barely recognized 20 years ago now affects at least 150,000 patients in the United States with three-quarters being adults. With greater appreciation for the emerging prevalence comes a plethora of epidemiologic questions for future investigators. Why do middle-aged men appear to be more affected than other populations? Why is EoE relatively uncommon among certain ethnic groups? Is there a possible birth cohort effect due to a yet undefined previous common exposure in these patients? What are the reasons for the regional variation of EoE prevalence? Accurate disease identification creates tremendous opportunities for collaborative research and effective therapeutic development across institutions. However, recognition of potential bias and promotion of the accurate use of the young ICD-9 code for EoE is essential so as to avoid the “greatest weakness” of epidemiologic research, that of “false-alarms and pseudo-epidemics.”15 ANDREW J. GAWRON, MD, PhD, MS Division of Gastroenterology and Hepatology and Center for Healthcare Studies Feinberg School of Medicine Northwestern University Chicago, Illinois
Clinical Gastroenterology and Hepatology Vol. 12, No. 4
IKUO HIRANO, MD Division of Gastroenterology and Hepatology Feinberg School of Medicine Northwestern University Chicago, Illinois
References 1. Shah A, Kagalwalla AF, Gonsalves N, et al. Histopathologic variability in children with eosinophilic esophagitis. Am J Gastroenterol 2009;104:716–721. 2. Sealock RJ, Kramer JR, Verstovsek G, et al. The prevalence of oesophageal eosinophilia and eosinophilic oesophagitis: a prospective study in unselected patients presenting to endoscopy. Aliment Pharmacol Ther 2013;37:825–832. 3. Hruz P, Straumann A, Bussmann C, et al. Escalating incidence of eosinophilic esophagitis: a 20-year prospective, populationbased study in Olten County, Switzerland. J Allergy Clin Immunol 2011;128:1349–1350. 4. Prasad GA, Alexander JA, Schleck CD, et al. Epidemiology of eosinophilic esophagitis over three decades in Olmsted County, Minnesota. Clin Gastroenterol Hepatol 2009;7:1055–1061. 5. Buckmeier B, Rothenberg M, Collins M. The incidence and prevalence of eosinophilic esophagitis. J Allergy Clin Immunol 2008;121:AB271. 6. Dellon ES, Jensen ET, Martin CF, et al. Prevalence of eosinophilic esophagitis in the United States. Clin Gastroenterol Hepatol 2014;12:589–596. 7. Rybnicek D, Hathorn K, Pfaff E, et al. Administrative coding is specific, but not sensitive, for identifying eosinophilic esophagitis. Dis Esoph 2013;in press. 8. Liacouras CA, Furuta GT, Hirano I, et al. Eosinophilic esophagitis: updated consensus recommendations for children and adults. J Allergy Clin Immunol 2011;128:3–20.e6; quiz 21–22. 9. Dellon ES, Gonsalves N, Hirano I, et al. ACG clinical guideline: evidenced based approach to the diagnosis and management of esophageal eosinophilia and eosinophilic esophagitis (EoE). Am J Gastroenterol 2013;108:679–692; quiz 693. 10. Kern EFO, Maney M, Miller DR, et al. Failure of ICD-9-CM codes to identify patients with comorbid chronic kidney disease in diabetes. Health Serv Res 2006;41:564–580. 11. Lopushinsky SR, Covarrubia KA, Rabeneck L, et al. Accuracy of administrative health data for the diagnosis of upper gastrointestinal diseases. Surg Endosc 2007;21:1733–1737. 12. Quach S, Blais C, Quan H. Administrative data have high variation in validity for recording heart failure. Can J Cardiol 2010; 26:306–312. 13. Chubak J, Pocobelli G, Weiss NS. Tradeoffs between accuracy measures for electronic health care data algorithms. J Clin Epidemiol 2012;65:343–349.e2. 14. 107th Congress. Rare Disease Act of 2002 Public Law 107–280. 15. Grimes DA, Schulz KF. False alarms and pseudo-epidemics: the limitations of observational epidemiology. Obstet Gynecol 2012; 120:920–927.
Conflicts of interest The authors disclose no conflicts. http://dx.doi.org/10.1016/j.cgh.2013.10.036
E
magnitude as previously published, single-center population studies.2–5 The estimate is likely more precise than prior studies based on the significantly larger sample size. Moreover, the insurance database was equally representative across regions of the country and age groups inclusive of both adults and children. In fact, one of the more notable findings reported is that in the time period studied, 75% of cases were adults. This reflects a paradigm shift in our thinking of EoE; EoE has returned to its origins as a predominantly adult disease. Although more precise, challenges certainly exist regarding the accuracy of a new ICD-9 code to define prevalent cases of a new disease. Esophageal eosinophilia is a histologic finding that also occurs in gastroesophageal reflux disease and proton-pump inhibitor–responsive eosinophilia. Other disease states such as eosinophilic gastroenteritis and connective tissue disorders, albeit rare, can be associated with esophageal eosinophilia. The actual diagnosis of EoE relies both on histopathology and clinical context as the disease definition continues to evolve.8,9 More specific biomarkers of the disease are being studied that may refine the diagnosis. Dellon and coauthors suggest that the ICD-9 code was specific for EoE; however, their validation was performed at a large tertiary referral center that is well known for treating esophageal disorders. Whether or not other practice settings are accurately classifying patients with EoE (and hence using the ICD-9 code correctly) is not known. Greater awareness of the existence of the ICD-9 code in an EoE tertiary center may overestimate the sensitivity whereas greater utilization of up-to-date diagnostic criteria in a tertiary center may overestimate the specificity of the code. In addition, challenges of using ICD-9 codes to define disease states has been documented in numerous instances of diseases with much longer histories than EoE including diabetes and gastroesophageal reflux disease.10–12 Patients may be assigned suspected diagnoses (ICD-9 codes) based on symptom presentation or endoscopic findings prior to actual confirmation of the disease. Another important limitation to the generalizability of the prevalence estimates in the current study pertains to the insurance database. The prevalence of EoE in the insured population may not accurately reflect the demographic differences in the uninsured. Furthermore, the database excluded patients older than age 65 and therefore cannot be extrapolated to the entire US population. Because of the relatively short existence of the EoE ICD code, there is significant opportunity among the gastroenterology, allergy, and pathology communities to ensure that it is used as accurately as possible. As the world of “big data” and use of electronic health records expands for research and clinical practice, data-mining techniques will allow for large population-based studies comparing treatments and monitoring patient outcomes across practice settings and regions. Incident user study designs are useful in limiting Clinical Gastroenterology and Hepatology 2014;12:597–598
598
Gawron and Hirano
bias in retrospective studies aimed at assessing comparative effectiveness of different therapies. As the therapeutic armamentarium for EoE increases and becomes more variable, it is important that researchers be able to accurately define cases and controls in retrospective epidemiologic and therapeutic comparison studies. The high reported specificity for the ICD-9 needs validation in cohorts outside of tertiary networks. Furthermore, the low sensitivity (37%) the authors quote from prior work validating the use of the ICD-9 code with their institutional database of confirmed EoE cases7 implies there are currently many false negatives. Thus, any future epidemiologic studies using this code should be very cautious in defining control groups as decreased sensitivity will not substantially bias relative risk as long as misclassification is nondifferential.13 The authors argue that because of the low sensitivity (ie, w37%) of using the ICD-9 code to identify EoE, the true prevalence of EoE could be as high 411,222 cases of EoE in the United States. EoE is currently considered a rare or “orphan” disease as defined by the Rare Disease Act of 2002 (disease affecting less than 200,000 individuals).14 This affords research opportunities via the Office of Rare Diseases Research and affiliated Rare Disease Clinical Research Network. Thus, accurate prevalence estimates have very important consequences for EoE patients, pharmaceutical companies, and researchers. At this time, we would caution against extrapolating population prevalence for EoE until longer-term data and further validation is available. In summary, Dellon and colleagues have provided a more precise and probably more accurate US population prevalence estimate of EoE. A disease that was barely recognized 20 years ago now affects at least 150,000 patients in the United States with three-quarters being adults. With greater appreciation for the emerging prevalence comes a plethora of epidemiologic questions for future investigators. Why do middle-aged men appear to be more affected than other populations? Why is EoE relatively uncommon among certain ethnic groups? Is there a possible birth cohort effect due to a yet undefined previous common exposure in these patients? What are the reasons for the regional variation of EoE prevalence? Accurate disease identification creates tremendous opportunities for collaborative research and effective therapeutic development across institutions. However, recognition of potential bias and promotion of the accurate use of the young ICD-9 code for EoE is essential so as to avoid the “greatest weakness” of epidemiologic research, that of “false-alarms and pseudo-epidemics.”15 ANDREW J. GAWRON, MD, PhD, MS Division of Gastroenterology and Hepatology and Center for Healthcare Studies Feinberg School of Medicine Northwestern University Chicago, Illinois
Clinical Gastroenterology and Hepatology Vol. 12, No. 4
IKUO HIRANO, MD Division of Gastroenterology and Hepatology Feinberg School of Medicine Northwestern University Chicago, Illinois
References 1. Shah A, Kagalwalla AF, Gonsalves N, et al. Histopathologic variability in children with eosinophilic esophagitis. Am J Gastroenterol 2009;104:716–721. 2. Sealock RJ, Kramer JR, Verstovsek G, et al. The prevalence of oesophageal eosinophilia and eosinophilic oesophagitis: a prospective study in unselected patients presenting to endoscopy. Aliment Pharmacol Ther 2013;37:825–832. 3. Hruz P, Straumann A, Bussmann C, et al. Escalating incidence of eosinophilic esophagitis: a 20-year prospective, populationbased study in Olten County, Switzerland. J Allergy Clin Immunol 2011;128:1349–1350. 4. Prasad GA, Alexander JA, Schleck CD, et al. Epidemiology of eosinophilic esophagitis over three decades in Olmsted County, Minnesota. Clin Gastroenterol Hepatol 2009;7:1055–1061. 5. Buckmeier B, Rothenberg M, Collins M. The incidence and prevalence of eosinophilic esophagitis. J Allergy Clin Immunol 2008;121:AB271. 6. Dellon ES, Jensen ET, Martin CF, et al. Prevalence of eosinophilic esophagitis in the United States. Clin Gastroenterol Hepatol 2014;12:589–596. 7. Rybnicek D, Hathorn K, Pfaff E, et al. Administrative coding is specific, but not sensitive, for identifying eosinophilic esophagitis. Dis Esoph 2013;in press. 8. Liacouras CA, Furuta GT, Hirano I, et al. Eosinophilic esophagitis: updated consensus recommendations for children and adults. J Allergy Clin Immunol 2011;128:3–20.e6; quiz 21–22. 9. Dellon ES, Gonsalves N, Hirano I, et al. ACG clinical guideline: evidenced based approach to the diagnosis and management of esophageal eosinophilia and eosinophilic esophagitis (EoE). Am J Gastroenterol 2013;108:679–692; quiz 693. 10. Kern EFO, Maney M, Miller DR, et al. Failure of ICD-9-CM codes to identify patients with comorbid chronic kidney disease in diabetes. Health Serv Res 2006;41:564–580. 11. Lopushinsky SR, Covarrubia KA, Rabeneck L, et al. Accuracy of administrative health data for the diagnosis of upper gastrointestinal diseases. Surg Endosc 2007;21:1733–1737. 12. Quach S, Blais C, Quan H. Administrative data have high variation in validity for recording heart failure. Can J Cardiol 2010; 26:306–312. 13. Chubak J, Pocobelli G, Weiss NS. Tradeoffs between accuracy measures for electronic health care data algorithms. J Clin Epidemiol 2012;65:343–349.e2. 14. 107th Congress. Rare Disease Act of 2002 Public Law 107–280. 15. Grimes DA, Schulz KF. False alarms and pseudo-epidemics: the limitations of observational epidemiology. Obstet Gynecol 2012; 120:920–927.
Conflicts of interest The authors disclose no conflicts. http://dx.doi.org/10.1016/j.cgh.2013.10.036
