Re: When and to What End Do Pathologists Agree?
Elliott Foucar · JNCI Journal of the National Cancer Institute · 1998
In a recent editorial discussing diagnostic precision in breast cancer, Page et al. ( 1 ) present an upbeat view of the troublesome area of precision in the diagnosis of ductal proliferative lesions. Unfortunately, their view is not supported by the literature or by the experience of other pathologists. Their editorial reports that the precision study by Schnitt et al. ( 2 ) showed that evaluation of ductal proliferative lesions resulted in a “greater than 90% agreement when the several pathologists involved used agreed-upon criteria” ( 1 ). However, the article by Schnitt et al. documented disagreement at the benign-malignant threshold in 33% of the cases, demonstrating that, although non-necrotizing ductal carcinoma in situ (DCIS) is an appealing theoretic step in the development of some cases of invasive carcinoma, this step remains refractory to histopathologic definition that would result in satisfactory diagnostic precision. Further evidence that this subtype of DCIS is a failed diagnostic category is provided by the findings in a study of diagnostic precision conducted by Rosai ( 3 ) and by the findings in a study by Wells et al. ( 4 ) that prompted the editorial by Page et al. In the Rosai study ( 3 ), there was disagreement at the benignmalignant threshold in 40% of the proliferative ductal cases, and in the study by Wells et al. ( 4 ), “critical disagreements occurred primarily in the differentiation between diagnoses of benign with atypia and noninvasive malignant.” More important, pathologists have found that the diagnosis of nonnecrotizing DCIS fails in daily practice. For example, a breast consultant at the Armed Forces Institute of Pathology ( 5 ), who was also a co-author of the article by Schnitt et al. cited by Page et al. noted, “…interobserver variability in separating AIDH [atypical intraductal hyperplasia] from DCIS remains notoriously high among pathologists.” Page et al. seek to avoid the problems with precision by advocating “central review,” but the studies by Schnitt et al. ( 2 ) and Rosai ( 3 ) document low precision among the very subspecialty pathologists who would be performing central review. Central review structures the diagnostic process by defining which pathologist is correct; however, the central review is a political solution to the inherent low precision of non-necrotizing DCIS, and it does not address the underlying defects in the diagnostic category. When theories of cancer development are used to declare a portion of a morphologic continuum to be a diagnostic category, pathologists are often faced with a lack of distinctive features, leading to poor diagnostic precision. Without precision, we lack one of the basic tools required for improving diagnostic accuracy. Unfortunately, diagnostic categories can often develop a political momentum that makes them difficult to modify or replace, even when they have deficiencies that compromise their clinical or scientific value. With 23 statistical editors and a neutral position on the issue, a publication like the Journal of the National Cancer Institute would appear to be well positioned to sponsor an evidence-based investigation of the validity of non-necrotizing DCIS as a malignant diagnostic category that is distinct from benign atypical hyperplasia. A good starting point for the evaluation would be: What do the studies by Schnitt et al., Rosai, and Wells et al. say about precision in the diagnostic setting? Would an objective reader conclude that these studies support DCIS as a valid diagnostic category, or do they provide objective evidence that this experiment in labeling many cytologically low-grade intraductal proliferations as “carcinoma” has been a failure?