超声诊断甲状腺结节证实偏倚校正的临床研究

Clinical study on corrected estimation of the verification bias on thyroid nodules diagnosed by ultrasonography

目的评估超声诊断甲状腺结节良恶性证实偏倚校正后的敏感性和特异性,为临床客观而全面地认识超声诊断的结果提供依据。方法采用最大似然估计值法对超声诊断甲状腺结节良恶性的敏感性和特异性的证实偏倚进行校正,并对其校正前后的95%置信区间进行对比。结果超声诊断恶性的患者接受手术证实的几率为35.10%(206/587),而超声诊断良性的患者接受手术证实的几率为11.31%(306/2705),两者差异有非常显著统计学意义(P<0.001)。超声诊断的未校正敏感性为87.67%(95%置信区间:82.34%～93.00%),校正后的敏感性下降到69.63%(95%置信区间:59.69%～79.57%),两者95%置信区间不存在交集,说明两者差异有显著统计学意义;未校正特异性为78.69%(95%置信区间:72.05%～85.33%),校正后的特异性提高到91.97%(95%置信区间:90.50%～93.44%),两者95%置信区间也不存在交集,说明两者的差异也具有显著统计学意义。经校正后,超声判断甲状腺结节良恶性的敏感性低于校正前,特异性却较校正前显著提高。结论超声检查发现甲状腺恶性病变的敏感性较低,但其排除恶性病变的特异性较强。

Objective To estimate the sensitivity and specificity of the benign and malignant lesions of the thyroid nodules diagnosed by ultrasonography （US） after corrected the verification bias, in order to supplying evidence for clinical practice to objective and overall reading the results of US. Methods The verification bias of the sensitivity and specificity of the benign and malignant lesions of the thyroid nodules diagnosed by US were corrected using the methods of maximum likelihood estimators, and compared the 95% confidence intervals between them. Results The operated verification rates of malignant patients diagnosed by US were 35.10% （206/587）, but on benign patients, it was 11.31% （306/2705）. There were significant difference between them （P〈0. 001）. The uncorrected sensitivity of US .was 87.67% （the 95 % confidence intervals were 82.34% - 93. 00%）, and the corrected sensitivity was decreased to 69.63% （the 95% confidence intervals were -59.69%-79.57%）. The 95% confidence intervals of them had not crossed each other, and had significant difference between them. The uncorrected specificity of US was 78. 69% （the 95% confidence intervals were 72.05% -85.33%）, and the corrected specificity was increased to 91.97 % （the 95 % eonfidenee intervals were 90.50 %-93.44 %）. The 95 % confidence intervals of them also had not crossed each other, and had significant difference between them in the same. After corrected the verification bias, the sensitivity of the benign and malignant lesions of the thyroid nodules diagnosed by US is decreased, and the specificity is increased. Conclusion The sensitivity of ultrasonography to detect malignant thyroid lesions is relative lower, but the specificity is relative higher.