Cellular adequacy for thyroid aspirates prepared by ThinPrep: how many cells are needed?

Although it is well established that ThinPrep introduces artifacts to thyroid aspirates, no criteria have been established for adequacy of such specimens. This study evaluates the adequate number of cells needed to establish the correct diagnosis based on ThinPrep alone. A total of 218 thyroid aspirates prepared by TP with surgical pathology follow-up were reviewed. The cellularity was calculated as follows: Count the total number of clusters, randomly select 10 clusters and count each, calculate the average number per cluster and multiply by the total number of clusters. A minimum number of 6 clusters with 10 cells each was arbitrary established to assume adequacy for a definitive diagnosis. Cytologic diagnoses were classified as: Nondiagnostic (ND), cystic contents, thyroiditis, nodular hyperplasia (NH), follicular/Hurthle (F/H) cell lesion, F/H cell neoplasm, and carcinoma: qualify. Histologic diagnoses were classified as: Cyst (colloid or otherwise), thyroiditis, NH, F/H adenoma, F/H carcinoma, carcinoma: qualify. Appropriate treatment triage was considered to be clinical for the former 4 cytologic categories and surgical for the latter 3 with ND warranting repeat aspiration. The results were subjected to logistic regressions analysis and contingency tables correlating the number of cells with the cytologic and histologic diagnosis as well as with treatment triage. Cellularity of sample was ranked in 10 deciles according to the number of cells and in 4 quartiles according to the number of clusters. The agreement percentage, for both diagnostic and treatment, was computed for each decile and quartile. 146 (67%) cases had cells and received a diagnosis while 72 (33%) were acellular. Of the 146 cases, 21 contained histiocytes or colloid only. 91/146 (62.3%) were correctly diagnosed and 123/146 (84.3%) would have been correctly triaged for treatment based upon the cytologic diagnosis. Samples with 180 cells or fewer had an agreement rate below 50%. Agreement rate increases to 80% when cellularity is 180-320. Above 320 agreement rate remains high but not uniformly. Total number of clusters did not play an independent role and only the number of cells per cluster had a significant correlation with diagnostic agreement. A 25-cell increase in average cells per cluster increases the odds of agreement between diagnoses by 65%. Copyright 2007 Wiley-Liss, Inc.