LIMITES DA TOMOGRAFIA COMPUTADORIZADA NA AVALIAÇÃO DO CARCINOMA ESPINOCELULAR DA LOJA TONSILAR LIMITATIONS OF COMPUTED TOMOGRAPHY ON THE ASSESSMENT OF THE SQUAMOUS CELL CARCINOMA OF THE TONSIL

Introduction: A comparative assessment between image interobserver analysis and clinical evaluation is mandatory for tonsilar neoplasms. Objective: The aim of the current study was to investigate the interobserver agreement of computed tomography scans for local and regional spread of the tonsillar fossa squamous cell carcinoma (SCC). Method: Twenty-five computed tomography images from 25 patients with SCC of the tonsillar fossa were selected and evaluated from 1990 to 1997. All patients were from the Service of Head and Neck Surgery, Hospital Heliopolis, Brazil. Twenty-five computed tomography examinations were reported retrospectively and independently by two experienced radiologists. They evaluated the involvement of the surrounding structures and the lymph nodes metastases. All patients were previously untreated by the time of the imaging examination. There was a marked predominance of men in our series, with 24 men and one woman; all but one had a long history of alcohol and tobacco use. The mean age was 56. SCC was confirmed through biopsy and histopathologic examination in all cases. Clinical data from medical charts were available to the readers. The proportion of agreement between the two observers was measured using the kappa statistics. Results: Good to excellent kappa scores (kappa, 0.61 to 1.0) was obtained in 8 out of 21 analyzed sites (nasopharynx, base of the tongue, hypopharynx, oral tongue/floor of the mouth, retromolar trigone, low jugular, accessory spinal, and retropharygeal lymph node levels). The rate of agreement was moderate in eight evaluated sites (kappa, 0.41 to 0.60); mastigator space, parapharyngeal space, posterior wall of the pharynx, mandibles branches, submandibular, high jugular, and middle jugular lymph nodes levels. A minimum agreement (kappa, 0.20 to 0.40) was obtained in three evaluated sites (soft palate, pre-epiglottic space and carotid space). The rate of agreement was insignificant (kappa ≤ 0.20) in the submentonian lymph node level. It was not possible to calculate the Kappa index for skull base. Conclusion: The proportion of acceptable item scores (good and excellent – kappa, 0.61 to 1.0) was obtained in 8 out of 19 analyzed sites. It was not acceptable (kappa < 0.61) in eleven. Kappa statistical was not possible to be performed in one case and it was insignificant in another one.