Blood metabolomics has been widely used for discovering potential metabolite biomarkers of various diseases. In this study, we report our investigation of the effects of freeze-thaw cycles (FTCs) of human serum samples on quantitative metabolomics using a differential chemical isotope labeling (CIL) LC-MS method. A total of 99 serum samples collected from healthy individuals (47 females and 52 males) were subjected to five FTCs, followed by 12C-/13C-dansylation labeling LC-MS analysis. A total of 2790 peak pairs or metabolites were relatively quantified among the 495 comparative samples, including 150 positively identified metabolites, 235 high-confident putatively identified metabolites and 1949 mass-matched metabolites from database searches. Multivariate analysis of the metabolome data showed a clustering of the third to fifth FTC samples in contrast to the separation of the first and second FTC samples, indicating that the extent of FTC-induced metabolome changes became smaller after the third cycle. The changing patterns among the FTC-effected metabolites were found to be complex. Using sex as a biological factor for grouping, we observed a clear separation of males and females when the samples were subjected to the same number of FTCs. However, when the male- and female-samples with different numbers of FTCs were compared, the number of significant metabolites found in male-female comparison increased dramatically, indicating that FTC effects could lead to a large number of false positives in biomarker discovery. Finally, we proposed a method of detecting the FTC effects by reanalyzing the original samples after subjecting them to an additional FTC.