Barrett’s esophagus is surveyed to prevent the development of esophageal adenocarcinoma. The end point of Barrett’s esophagus surveillance is the establishment of low- or high-grade dysplasia, carcinoma in situ, or invasive carcinoma because these diagnoses lead to additional therapeutic interventions. It would seem evident that a process such as dysplasia, which has been established in pathology almost since its inception, would be well defined and the criteria well delineated. However, a seminal article on Barrett’s dysplasia showed that 12 leading GI pathologists from centers of excellence in Barrett’s esophagus showed a large interobserver variability for all non–high-grade dysplasia diagnoses even after there was a consensus conference as to what morphologic changes should be categorized as dysplasia. The results showed a 90% agreement if two diagnostic categories were used (negative or high-grade dysplasia vs. high-grade dysplasia or more significant lesion) and a 75% agreement if three diagnostic categories were used (BE, negative for dysplasia, BE indefinite for dysplasia or low-grade dysplasia, and BE with high-grade dysplasia or more significant lesion).
The authors also showed a "barbell distribution" of agreement in that there was fairly good agreement at the low and high end of the spectrum (negative for dysplasia and high-grade dysplasia) and very poor agreement in the middle ground between indefinite for dysplasia and low-grade dysplasia. Even so, they never reached a level of greater than 90% agreement.5
This inherent subjectivity to morphologic analysis means there is no gold standard. If no gold standard exists, then it is evident that comparison of diagnostic acumen between pathologists and institutions is meritless. These limitations have led pathologists to test various biomarkers or ancillary studies to help and increase accuracy and precision and to render reproducible diagnoses between pathologists. It is universally accepted that high-grade dysplasia is the best predictor of progression in Barrett’s esophagus to adenocarcinoma. However, the practical application of this standard is limited by the subjective nature of the criteria for establishing high-grade dysplasia and that some reactive conditions cannot be readily distinguished from high-grade dysplasia by histological assessment alone. Therefore, pathologists use biomarkers to distinguish dysplasia from its histological mimics and to help and grade the level of dysplasia.6
The problem of interobserver variability is not limited to Barrett’s esophagus but extends to colon cancer screening. For colon cancer screening, risk stratification is based on the size and morphology of the lesion. The gastroenterologist determines the size of the lesion at colonoscopy, and the pathologist then defines the morphology. There is no consensus among pathologists as to what constitutes high-grade dysplasia and villous morphology, and the endoscopic estimate of polyp size is unreliable.7,8
The reason for errors and disagreement is the inherently subjective nature of pathology. The best way to resolve this dilemma is to bring more objective standards to bear and to establish a gold standard. One way of objectifying a subjective interpretation is to submit it to a binary "yes"/"no" test. This is the role of add-on studies.
Liability
The pathologist is a physician subject to litigation as are all health care providers. The pathologist’s clinical counterpart shares the burden of the missed pathologic diagnosis. The use of ancillary studies increases the likelihood of a correct diagnosis and minimizes the likelihood of the missed or wrong diagnosis.
Conclusion
In the end, the eye sees what the mind knows, but what the mind knows and the eye sees are highly variable among many pathologists; and in some cases the result is a nonreproducible diagnosis. Clearly, this is not an acceptable state. If a diagnosis cannot be reproducibly rendered between pathologists in the same institution or between collaborative institutions, any study that relies on such diagnosis is inherently flawed. The pathologist, the clinician, and the patient should expect and demand the "right" or at least the "best" answer. However, this lack of reproducibility means that the patient who has their slides sent to a large number of institutions seeking the "right diagnosis" is on a fool’s errand. The objective information bought by ancillary studies achieves this goal in some cases. The failure to use ancillary studies fully may do the patient and the clinical community a gross disservice.
The GAO report showing possible overuse related to conflict of interest has initiated considerable interest, scrutiny, and conversation in Congress and CMS.1
Recent reductions in the reimbursement for 88305 have tempered the need for further cost reductions (at the expense of pathologists and owners of laboratory services), but we can expect Congress to revisit this issue in the future. Our best defense for preservation of this Stark exemption is to practice evidence-based medicine where we can defend our decisions on the basis of patient needs and improved health outcomes.
