Role of frozen section, sex, age and tumour size in differentiating follicular adenoma from carcinoma: a meta-analysis

Abstract

Abstract Background Frozen section (FS) has been the mainstay of intraoperative decision-making in thyroid neoplasms for many years. However, since the widespread use of preoperative fine-needle aspiration, the need for FS in patients with follicular neoplasms (FNs) has been debated. Preoperative features may be better at distinguishing follicular carcinoma (FC) from adenoma (FA). One reason why these issues are unresolved is because clinical studies are often too small to reach statistical significance. The purpose of this study was to combine modern trials and use the technique of meta-analysis to determine the efficacy of FS and the role of sex, age and tumour size in patients with FNs. Methods Inclusion criteria were published and unpublished studies between 1990 and 1999 inclusive, in which patients had a permanent pathology (PP) diagnosis of FA or FC (including Hurthle cell tumours) and underwent FS or had clinical features recorded. A pooled sample of the combined data was analysed for the prognostic variables and outcomes. P < 0·05 was considered significant. Results Nineteen studies were included. FS was evaluated in 11 studies (n = 2204). FS matched PP in 82 per cent of patients and was considered helpful. FS was indeterminate in 13 per cent of cases and was considered unhelpful. FS was detrimental in 4 per cent of patients. In 27 per cent of FCs, FS reported FA (false negative) and in 1 per cent of FAs, FS reported FC (false positive). Overall, FS had an 87 per cent sensitivity, 48 per cent specificity, 92 per cent positive predictive value, 35 per cent negative predictive value and 82 per cent accuracy. Clinical features were examined in ten studies (n = 1980). Nine studies (n = 1780) reported sex; 20 per cent of patients were male. A disproportionately high percentage of patients with FC were male (28 per cent) compared with female (P < 0·001). Two studies (n = 548) reported age; 36 per cent of patients were over 50 years old. A disproportionately high percentage of patients with FC were aged over 50 years (52 per cent) compared with those younger than 50 years (P < 0·001). Six studies (n = 1620) reported tumour size; 19 per cent of FNs were larger than 3–5 cm. A disproportionately high percentage of FCs were larger than 3–5 cm (33 per cent) compared with smaller lesions (P < 0·001). Conclusion FS is able to differentiate FC from FA in over 80 per cent of patients. It has a false-positive rate of less than 1 per cent and therefore may be trusted most when it reports a FC. In cases in which FS reports FA, male sex, age greater than 50 years and tumour size larger than 3–5 cm can be used as indicators that a FN will be a FC.