Raman spectroscopy combined with multivariate analysis was used to examine the molecular mechanism of carboplatin toxicity in human nasopharyngeal carcinoma epithelioid cell line CNE-1. Multivariate analysis was performed through principal component analysis combined with linear discriminant analysis, through which a diagnostic algorithm that could discriminate between the control and treatment groups was generated. Compared with the control group (0 μmol/L, 0 hour), the 2 protein Raman peaks (1002 cm−1 and 1659 cm−1) and the 2 deoxyribonucleic acid Raman peaks (1303 cm−1 and 1338 cm−1) in the treatment group decreased with increasing carboplatin concentrations as well as with increasing treatment durations. Using principal component analysis combined with linear discriminant analysis, the discrimination accuracy, sensitivity, and specificity between the control group and the cell groups treated at 12 and 16 μmol/L carboplatin for 24 hours were all 100%, suggesting that 16 μmol/L is suitable as the therapeutic dosing. Using the same method, the discrimination accuracy, sensitivity, and specificity between the control group and the cell groups that were treated with 12 μmol/L carboplatin for 24 and 36 hours were all 100%, indicating that 24 hours could be suitable treatment duration. These results suggest that Raman spectroscopy combined with multivariate statistical analysis could be a useful tool for assessing carboplatin-induced cytotoxicity in human nasopharyngeal carcinoma cells.