Optical spectroscopy (INVOS) is unreliable in detecting breast cancer

Recently a new nonionizing, nonimaging technique for evaluating risk of breast cancer, in vivo optical spectroscopy (INVOS), became available. The procedure evaluates the biochemical composition of the breast with spectrophotometry to provide a risk number related to the development of breast cancer. INVOS was evaluated in a prospective study of 180 women who were referred to our institution for needle localization and excisional biopsy of 181 breast lesions. All women had mammography and INVOS before surgical biopsy. The INVOS risk numbers and mammographic findings were compared with the pathologic diagnoses. Of 39 breast cancers, 19 (49%) had high-risk INVOS numbers, as did 52% of women with benign disease. INVOS had a sensitivity of 95%, specificity of 4%, and a positive predictive value of 21% for the detection of breast cancer. Mammography predicted malignancy with a sensitivity of 97%, specificity of 48%, and a positive predictive value of 34%. Receiver-operating-characteristic (ROC) curve analysis also was performed. The ROC area for mammography was 0.85 compared with 0.43 for INVOS, indicating a significantly greater (p less than .0000001) ability for mammography to detect breast cancer. We conclude that because INVOS cannot accurately predict the presence of breast cancer, it is unlikely that INVOS can be used to predict which patients are at risk for the development of breast cancer.