Polybrominated diphenyl ethers (PBDEs) are used as additive flame retardants. Because they lack the ability to form chemical bonds, PBDEs can easily enter the sediment environment. The accurate qualitative and quantitative analysis of PBDEs in sediments is of great importance for the accurate assessment of PBDE pollution in this environment. Sediments contain many impurities. Therefore, PBDEs in sediment should be purified before analysis to reduce the matrix effect. A method based on gas chromatography-negative chemical ionization-mass spectrometry (GC-NCI/MS) was developed to determine 13 PBDEs in marine sediment samples using a column packed with deactivated silica gel, acidified silica gel, Florisil, and anhydrous sodium sulfate. Sediment samples were extracted by ultrasonication with a mixed solvent of n-hexane-dichloromethane (3∶1, v/v). After two cycles of ultrasonic extraction, the extract was purified by a composite chromatographic column and eluted with n-hexane-dichloromethane (3∶1, v/v). Thirteen PBDEs were determined by GC-NCI/MS in selected-ion monitoring (SIM) mode. The effects of different fillers, eluents, and elution volumes on the purification of PBDEs in the composite column were compared and analyzed, and the GC-NCI/MS analysis conditions were optimized. Three different packing columns were used to purify the sample extract. The first column was packed with 3 g of deactivated silica, 6 g of acidic silica, 3 g of deactivated silica, 3 g of Florisil, and 6 g of anhydrous sodium sulfate; the second column was packed with 3 g of Florisil, 3 g of deactivated silica, 6 g of acidic silica, 3 g of deactivated silica, and 6 g of anhydrous sodium sulfate; and the third column was packed with 3 g of deactivated silica, 6 g of acidified silica, 3 g of deactivated silica, and 6 g of anhydrous sodium sulfate. Among these columns, that packed with 3 g of deactivated silica, 6 g of acidic silica, 3 g of deactivated silica, 3 g of Florisil, and 6 g of anhydrous sodium sulfate showed the best purification effect. The 13 PBDEs showed good linearity in the mass concentration range of 0.1-20 μg/L with correlation coefficients (r2) greater than 0.995 (decabromodiphenyl oxide (BDE-209), r2>0.99). The limits of quantification (S/N=10) was 0.002-0.126 μg/kg. The average recoveries of the 13 PBDEs at three spiked levels of 0.2, 1.0, and 4.0 μg/kg were 85.3%-101.3%, 84.8%-113.6%, and 86.3%-94.7% with relative standard deviations of 4.4%-14.0%, 0.4%-4.9%, and 1.9%-6.6%, respectively. These findings indicate that the method has high sensitivity and accuracy as well as good precision. Finally, the method was applied to the analysis and detection of PBDEs in actual marine sediment samples. The results revealed that the sediment samples contained different contents of the 13 PBDEs, and high detection rates were obtained for lower-brominated PBDE homologs. The detection rate of bis(4-bromophenyl) ether (BDE-15) was 100%, and the detected content of BDE-209 was as high as 60.49 μg/kg. These results demonstrate that the developed method is suitable for the accurate qualitative and quantitative analysis of PBDEs in marine sediment samples.