TOPIC: Thyroid cancer
Title: Inter-observer and intra-observer reproducibility in the histo-pathology of follicular thyroid carcinoma.
Authors: Franc B, De La Salmoniere P, Lange F, Hoang C, Louvel A, De Roquancourt A, Vilde F, Hejblum G, Chevret S, & Chastang C.
Reference: Human Pathology 34: 1092-1100, 2003
Thyroid nodules are very common, and most are benign. A fraction of them is malignant. Until now, the diagnosis rests on pathological examination. Papillary thyroid carcinomas (PTC) are relatively easy to diagnose on the basis of their nuclear features and the characteristic papillary form of their structure. Follicular thyroid carcinomas (FTC) and sometimes the follicular variants of PTCs are much more difficult to diagnose. These difficulties (as well as precautionary attitudes) may therefore lead to over-diagnosis. The reproducibility of these diagnoses has rarely been tested.
To evaluate inter- and intra-observer reproducibility in the histopathology of follicular carcinoma (FTC).
Design, Setting and Patients
Forty-one anonymous FTC pathology slides were reviewed independently by five established specialized pathologists and 31 of them were also evaluated twice by the same pathologist. A “final consensus diagnosis” (FCD) was made at the end of the study. Agreement was evaluated by statistical methods.
The agreement between the 5 observers’ initial diagnosis and the FCD was 0.69, 0.41, 0.35, 0.28 & 0.11, respectively. This strongly suggests a leadership phenomenon. The FCD classified 30 of the cases as malignant, including 24 cases diagnosed as FTC and 11 as adenomas. There was unanimous agreement for 13 of the 24 FTCs. Discrepancies occurred in 4/7 cases classified as minimally invasive FTC by the FCD. Inter-observer and intra-observer agreement for FTC diagnosis was 0.23 and 0.68 respectively.
This study showed that the diagnostic reproducibility of minimally invasive FTC is low.
The correct diagnosis of a tumor is essential for the clinician since it has prognostic and therapeutic consequences. Although pathologists are well aware of the subjectivity of their diagnoses, they rarely emphasize this point and clinicians often take their diagnosis at face value. An article such as this one is therefore illuminating, but as it questions the validity of a universally used tool it must have been very difficult to publish and very easy to forget. There are inconvenient truths that are best left ignored!
While the diagnosis of classical papillary and anaplastic thyroid carcinomas is relatively clear-cut, distinction between follicular adenoma and minimally invasive or initial stage follicular carcinoma remains difficult. First, there is the problem of the representativity of the samples examined. Second, an in situ follicular adenoma is a follicular adenoma until it becomes invasive; then, this invasion needs to be detected. The ‘would be’ criminal is only convicted when his crime has been committed and shown!
A few studies on the reproducibility of pathological diagnosis in the field of thyroid carcinoma have been published between 1976 and 1993, and another study on a few cases was published in 2002. The present study by Brigitte Franc and her colleagues considers diagnoses made on the same sets of slides by specialized pathologists. Even in such privileged conditions, the reproducibility of the diagnoses by the same pathologist was bad (2 out of 3) and the reproducibility between different observers was dismal (1 out of 5). This is also true for the diagnosis of vascular invasion and nucleus optical clearing. These straightforward facts all point in the same direction. Considering that sampling hazards are not taken into account, such results indicate that clinicians should not take pathological diagnoses as definitive answers, at least in difficult cases. Obviously, this is even more strongly true in the diagnoses made by non-specialized histo-pathologists. Not everybody has access to the vast experience of Dillwyn Williams!
Similar mental reservations apply to diagnoses made on fine needle aspiration biopsies (FNAB), not withstanding the fact that FNAB provides no information about potential vascular invasion. A similar study on FNAB reproducibility would obviously be welcome.
One issue directly following this situation may be the tendency of the diagnostician to practice defensive (or cautious) medicine and, hence, always leave open the possibility of the presence of a cancer! Finally, this study shows the need for objective quantitative diagnostic tests that could result from present work, based for instance on the molecular biology of cancers (i.e. gene expression, etc.).
NB ( note from the Editor ): The selection for a personal commentary by Prof J. Dumont of an article published some years ago does not fall into the overall editorial direction of the Section ‘Thyroid News’ in Thyroid Disease Manager. However, in view of its ‘forgotten’ importance, we decided to accept his proposal to comment on the present article.
Summary and commentary prepared by Jacques Dumont (Related to Chapter 18 of TDM)