It takes two to tango: dance steps for diagnosing Barrett's esophagus

In the midst of a busy day of endoscopy, you read that the next patient on your schedule is a young woman referred by her primary care provider for open access for surveillance of Barrett's esophagus. The worried patient tells you that she had an upper endoscopy a few years previously at another facility and was told she had “precancer” in her esophagus. The endoscopy report is not available, and the patient does not recall whether any biopsy specimens were obtained. On your careful endoscopic examination, you find a small sliding hiatal hernia but no displacement of the squamocolumnar junction into the esophagus. In the recovery room, when you inform her that she does not have Barrett's esophagus and that she is likely at no increased risk for esophageal cancer, she is disbelieving and confused, which is of course exacerbated by the fog of her sedation.

Even if you do not allow open access upper endoscopy in your practice, you have likely encountered a similar scenario in your outpatient visits. In this month's Gastrointestinal Endoscopy, Corley et al1 have shown that 31% of patients assigned a billing diagnosis code for Barrett's esophagus do not in fact have Barrett's esophagus. This poses problems for patients, clinicians, and researchers. Furthermore, it is a symptom of the embarrassing inefficiency in our health care system. To improve on this will require endoscopists and pathologists to work together.

In their study, Corley et al manually reviewed the endoscopy and pathology records from 2470 patients served by Kaiser Permanente Northern California who had received either an International Classification of Disease, 9th revision (ICD-9) billing code for Barrett's esophagus, a Systematized Nomenclature of Medicine (SNOMED) pathology code for Barrett's esophagus, or both codes. Of those 2470 patients with a billing or pathology code of Barrett's esophagus, a single referral pathologist reviewed the pathology slides in 616 serially diagnosed patients. The referral pathologist then rereviewed a subset of 44 slides in a blinded fashion.

A few findings are particularly worth highlighting. Foremost is that at least 31% of those patients carrying an ICD-9 code for Barrett's esophagus ([390 + 128]/[571 + 1101]) do not actually have Barrett's esophagus on manual review of their records. A major reason is that approximately 50% of these patients do not have intestinal metaplasia identified histologically. Of those assigned a SNOMED pathology code for Barrett's esophagus, 24% were found not to have Barrett's esophagus ([330 + 128]/[798 + 1101]), in large part because of lack of endoscopic findings of Barrett's esophagus. In the United States, the accepted standard for the diagnosis of Barrett's esophagus requires both the endoscopic identification of columnar-appearing mucosa residing in the distal esophagus and the histologic identification of intestinal metaplasia in biopsy specimens obtained from that area—it takes two to tango.2, 3, 4

Overall, of those who were ultimately determined not to have Barrett's esophagus, 25% had a hiatal hernia, and 26% had esophagitis—both conditions may be confused with Barrett's esophagus. Wide tongues or confluent erosions can be confused as columnar mucosa. Likewise, if the endoscopist does not patiently wait through the respiratory cycle and examine the gastroesophageal junction with various levels of insufflation, a hiatal hernia (both small and large) can easily be mistaken as a columnar-lined esophagus rather than columnar-lined stomach. Biopsy specimens from this area may demonstrate intestinal metaplasia, but intestinal metaplasia of the gastric cardia does not represent the same risk for cancer as that of intestinal metaplasia of the esophagus.2

In addition to the reliability of diagnostic codes for Barrett's esophagus, Corley et al evaluated the reliability of local pathologists for diagnosing Barrett's esophagus. They found that the interpretation by local pathologists was very sensitive for Barrett's esophagus—only 1% of patients had their diagnosis changed by the referral pathologist from no Barrett's esophagus to having Barrett's esophagus. On the other hand, the local interpretation may only be moderately specific because 9.4% of those initially diagnosed with Barrett's esophagus were interpreted by the referral pathologist as not harboring Barrett's epithelium. However, this apparent inaccuracy of the local pathologists may simply be due to random errors expected of any pathologist. Even the referral pathologist in this study only had good, but not excellent, agreement with his own diagnosis made 2 months previously, changing his diagnosis 11% of the time. A limitation of this study is that it truly only judged the local pathologists against this single referral pathologist, and other expert Barrett's pathologists may claim they have greater internal consistency in their diagnoses.

The use of billing databases for research on Barrett's esophagus had long been hampered by the particular code used for Barrett's esophagus. Before 2003, the ICD-9 code for Barrett's esophagus was 530.2, which was also used for “ulcer of the esophagus,” including peptic, fungal, and pill induced. The code for Barrett's esophagus was subsequently changed to 530.85, which does not include other diagnoses. According to the authors, the code 530.2 had been used exclusively for Barrett's esophagus during the study period at Kaiser, but it is unlikely to have been used so specifically in most other settings. The assignment of a unique ICD-9 code was an important step toward clarifying the diagnosis of Barrett's esophagus for research purposes, but as Corley et al have shown, it by no means was a panacea. Research on Barrett's esophagus using administrative data will likely continue to require manual review of medical records to confirm the diagnosis of Barrett's esophagus.

The misuse of billing codes for Barrett's esophagus is likely in part due to the confusion among providers regarding what criteria are required to make the diagnosis of Barrett's esophagus, leading to overdiagnosis of Barrett's esophagus. A recent survey of endoscopists in the United Kingdom found that at least 28% did not believe that intestinal metaplasia was necessary to make the diagnosis,5 although this could be influenced by the differing guidelines in the United Kingdom on this topic.6 I am unaware of any studies among community endoscopists in the United States regarding their knowledge and acceptance of the recommended requirement of histologic evidence of specialized intestinal epithelium for the diagnosis of Barrett's esophagus. Given primary care provider confusion regarding whether Helicobacter pylori causes gastroesophageal reflux and whether it is best to perform screening for Barrett's esophagus while the patient is on or off proton pump inhibitors,7 it may be reasonable to assume that many primary care providers are unaware of the intricacies of making the diagnosis of Barrett's esophagus. Therefore, it may be too much to expect of a primary care provider overwhelmed with managing hypertension, diabetes, every ache and pain, and scheduling timely screening examinations to also referee between the endoscopist and the pathologist, particularly if their reports are poorly written. If the endoscopy report merely states “Barrett's esophagus” without any further explanation and the pathology report merely states “gastric epithelium,” it is easy to imagine how the primary care provider might mistakenly include Barrett's esophagus in the patient's problem list. Even if the patient is subsequently seen in follow-up by the endoscopist, who dictates a letter stating that the patient does not have Barrett's esophagus, once an error enters the medical record, even if subsequently corrected, it tends to creep back into the record. Such errors in the medical record can lead to inappropriate referrals for surveillance.

Repeated endoscopies on patients who are not at risk for cancer provides no benefit to the patient, while exposing them to risk for procedural complications, and ultimately raises the cost of medical care paid by employers, medical insurance policy holders, and taxpayers. Furthermore, an erroneous diagnosis of Barrett's esophagus could raise the patient's costs for life and medical insurance and possibly exclude patients from being offered medical insurance at all.8 Improving the efficiency of our surveillance program for esophageal adenocarcinoma will in part require a careful dance of endoscopists with pathologists, so here are a few simple steps for dancing the Barrett:

As endoscopists, many of us may be terrible dancers, but we are rather nimble with a scope. Extending that skillfulness to the documentation and communication regarding the diagnosis of Barrett's esophagus could greatly improve the efficiency of screening for esophageal adenocarcinoma.