本研究探討以高效能液相層析(HPLC)結合感應耦合電漿質譜法(ICP/MS)進行勞工職業性砷暴露尿液檢體生物偵測砷物種分析方法之可行性。研究中使用PRP-X100陰離子交換樹脂管與30mM NaH2PO4, pH 6.0移動相，在1.5mL/min的流洗流速下，As(Ⅲ)、DMA、MMA及As(Ⅴ)四物種層析滯留時間分別為1.67、2.30、2.94與5.10min。對於尿液樣品檢量線範圍10-100μg/L，以上四物種的檢量線線性相關係數均大於0.999。As(Ⅲ)、DMA、MMA及As(Ⅴ)四物種尿液樣品偵測極限分別為3.9、1.0、0.6與1.3μg/L。常人尿液檢體在20, 40與80μg/L砷標準溶液添加測試中，As(Ⅲ)、dimethylarsenate (DMA)、monomethylarsenate (MMA)及As(Ⅴ)四物種在三種不同濃度樣品平均回收率為89.0、104.8、109.7與124.5%，而在不同濃度樣品中，各物種最高與最低回收率之間差距則分別為9.2、2.4、2.5與8.5%。在四種尿液參考標準樣品測試分析方法的準確度中，除一低濃度樣品LypHochek®樣品376之四物種砷總和測定含量38.3μg/L比參考值29μg/L高出32%，落在參考值95%信賴區間(23-34μg/L)之外，另外一種低濃度與兩種高濃度標準品(2524, 377, 2525)測定值81.7、256.6與170.4μg/L皆有在95%信賴區間52-86μg/L、183-274μg/L與167-201μg/L之內。

This study involved the evaluation of a method of arsenic speciation for urine samples using HPLC-ICP/MS in the biological monitoring of occupational exposure of workers. The retention times for As (Ⅲ), MMA, DMA, and As (Ⅴ) were 1.67, 2.30, 2.94, and 5.10 min, respectively, at a 1.5 mL/min flow rate of 30 mM NaH2PO4 with a PRP-X100 anion exchange column. At a calibration range of 10-100 μg/L, the correlation coefficients for all four species were above 0.999. The detection limits (DL's) of As (Ⅲ), MMA, DMA, and As (Ⅴ) in urine samples were 3.9, 1.0, 0.6 and 1.3 μg/L, respectively. The mean recoveries of spiked urine samples were 89.0, 104.8, 109.7 and 124.5% for the four arsenic species, with test concentrations of 20, 40, and 80 μg/L for each species. The coefficients of variance in the recovery rate of different levels were 9.2, 2.4, 2.5, and 8.5% for As (Ⅲ), MMA, DMA, and As (Ⅴ), respectively. Four certified urine reference samples were used to test the accuracy of this method. For the low-arsenic level sample, LypHochek® 376, the measured total arsenic species concentration was 38.3 μg/L, which was 32% higher than the reference value of 29 μg/L and fell outside the 95% confidence interval of 23-34 μg/L. For all three of the other samples, one low (2524) and two high (377 and 2525), the measured arsenic concentrations were in fair agreement with the reference values.