The accuracy of the SONOBREAST statistical model in comparison to BI-RADS for the prediction of malignancy in solid breast nodules detected at ultrasonography

详细信息    查看全文

• 作者：Regis R. Paulinelli ; ^{rrpaulinelli@gmail.com" class="auth_mail" title="E-mail the corresponding author} ; Luis-Fernando P. Oliveira ^{luisfernandopaduaoliveira@gmail.com" class="auth_mail" title="E-mail the corresponding author} ; Ruffo Freitas-Junior ^{ruffojr@terra.com.br" class="auth_mail" title="E-mail the corresponding author} ; Leonardo R. Soares ^{ribeiroufg@hotmail.com" class="auth_mail" title="E-mail the corresponding author}
• 关键词：Breast cancer ; Ultrasonography ; Diagnosis ; Sensitivity and specificity ; SONOBREAST
• 刊名：European Journal of Obstetrics & Gynecology and Reproductive Biology
• 出版年：2016
• 出版时间：January 2016
• 年：2016
• 卷：196
• 期：Complete
• 页码：1-5
• 全文大小：1362 K

文摘

The objective of the present study was to compare the accuracy of SONOBREAST for the prediction of malignancy in solid breast nodules detected at ultrasonography with that of the BI-RADS system and to assess the agreement between these two methods.

Study design

This prospective study included 274 women and evaluated 500 breast nodules detected at ultrasonography. The probability of malignancy was calculated based on the SONOBREAST model, available at www.sonobreast.com.br, and on the BI-RADS system, with results being compared with the anatomopathology report.

Results

The lesions were considered suspect in 171 cases (34.20%), according to both SONOBREAST and BI-RADS. Agreement between the methods was perfect, as shown by a Kappa coefficient of 1 (p < 0.001). SONOBREAST and BI-RADS proved identical insofar as sensitivity (95.40%), specificity (78.69%), positive predictive value (48.54%), negative predictive value (98.78%) and accuracy (81.60%) are concerned. With respect to the categorical variables (BI-RADS categories 3, 4 and 5), the area under the receiver operating characteristic (ROC) curve was 94.41 for SONOBREAST (range 92.20–96.62) and 89.99 for BI-RADS (range 86.60–93.37).

Conclusions

The accuracy of the SONOBREAST model is identical to that found with BI-RADS when the same parameters are used with respect to the cut-off point at which malignancy is suspected. Regarding the continuous probability of malignancy with BI-RADS categories 3, 4 and 5, SONOBREAST permits a more precise and individualized evaluation.