High mortality rates associated with oral cancers can be primarily attributed to the failure of current histological procedures in predicting recurrence. Identifying recurrence related factors can lead to improved prognosis, optimized treatment and enhanced overall outcomes. Serum Raman spectroscopy has previously shown potential in the diagnosis of cancers, such as head and neck, cervix, breast, oral cancers, and also in predicting treatment response. In the present study, serum was collected from 22 oral cancer subjects [with recurrence (n = 10) and no-recurrence (n = 12)] before and after surgery and spectra were acquired using a Raman microprobe coupled with a 40× objective. Spectral acquisition parameters were as follows: λex = 785 nm, laser power = 30 mW, integration time: 12 s and averages: 3. Data was analyzed in a patient-wise approach using unsupervised PCA and supervised PC-LDA, followed by LOOCV. PCA and PC-LDA findings suggest that recurrent and non-recurrent cases cannot be classified in before surgery serum samples; an average classification efficiency of ∼78% was obtained in after-surgery samples. Mean and difference spectra and PCA loadings indicate that DNA and protein markers may be potential spectral markers for recurrence. RS of post surgery serum samples may have the potential to predict the probability of recurrence in clinics, after prospective large-scale validation.

Serum Raman spectroscopy (RS) has previously shown potential in oral cancer diagnosis and recurrence prediction. To evaluate the potential of serum RS in oral cancer screening, premalignant and cancer-specific detection was explored in the present study using 328 subjects belonging to healthy controls, premalignant, disease controls, and oral cancer groups. Spectra were acquired using a Raman microprobe. Spectral findings suggest changes in amino acids, lipids, protein, DNA, and β-carotene across the groups. A patient-wise approach was employed for data analysis using principal component linear discriminant analysis. In the first step, the classification among premalignant, disease control (nonoral cancer), oral cancer, and normal samples was evaluated in binary classification models. Thereafter, two screening-friendly classification approaches were explored to further evaluate the clinical utility of serum RS: a single four-group model and normal versus abnormal followed by determining the type of abnormality model. Results demonstrate the feasibility of premalignant and specific cancer detection. The normal versus abnormal model yields better sensitivity and specificity rates of 64 and 80%; these rates are comparable to standard screening approaches. Prospectively, as the current screening procedure of visual inspection is useful mainly for high-risk populations, serum RS may serve as a useful adjunct for early and specific detection of oral precancers and cancer.