Thirteen studies (1,020 participants, range 12 to 299) were included in the review. Two studies included some patients with a primary tumour other than papillary thyroid carcinoma (other head and neck cancers). Most studies did not report the time interval between the index test and reference standard and only around half of the studies reported interpretation of the reference standard blinded to the results of the index test.
Pooled estimates of per patient sensitivity and specificity of real-time ultrasonography were 72% (95% CI 46% to 88%) and 98% (95% CI 84% to 100%) based on six studies with significant between-study heterogeneity. The median prevalence of cervical lymph node metastasis in patients with papillary thyroid carcinoma was 45% and linear regression suggested that sensitivity was not correlated with prevalence.
Pooled estimates of per lesion sensitivity and specificity of real-time ultrasonography were 63% (95% CI 47% to 76%) and 93% (95% CI 73% to 99%) based on seven studies with significant between-study heterogeneity. The median prevalence of cervical lymph node metastasis in patients with papillary thyroid carcinoma was 58%. Linear regression suggested that sensitivity was not correlated with prevalence.
There was no evidence of threshold effect. Meta-regression analysis, using individual QUADAS criteria as indicator variables, suggested that representativeness of the participant sample, time delay between index test and reference standard and blinded interpretation of the reference standard significantly affected estimates of test performance. The results of subgroup analyses were reported in the article. No data that compared real-time ultrasonography to other imaging modalities were reported.
There was no evidence of publication bias.