Urine and blood samples of cancer patients, treated with high doses of arsenic trioxide were analysed for arsenic species using HPLC-HGAFS and, in some cases, HPLC-ICPMS. Total arsenic was determined with either flow injection-HGAFS in urine or radiochemical neutron activation analysis in blood fractions (in serum/plasma, blood cells). The total arsenic concentrations (during prolonged, daily/weekly arsenic trioxide therapy) were in the microg mL(-1) range for urine and in the ng g(-1) range for blood fractions. The main arsenic species found in urine were As(III), MA and DMA and in blood As(V), MA and DMA. With proper sample preparation and storage of urine (no preservation agents/storage in liquid nitrogen) no analytical artefacts were observed and absence of significant amounts of alleged trivalent metabolites was proven. On the contrary, in blood samples a certain amount of arsenic can get lost in the speciation procedure what was especially noticeable for the blood cells although also plasma/serum gave rise to some disappearance of arsenic. The latter losses may be attributed to precipitation of As(III)-containing proteins/peptides during the methanol/water extraction procedure whereas the former losses were due to loss of specific As(III)-complexing proteins/peptides (e.g. cysteine, metallothionein, reduced GSH, ferritin) on the column (Hamilton PRP-X100) during the separation procedure. Contemporary analytical protocols are not able to completely avoid artefacts due to losses from the sampling to the detection stage so that it is recommended to be careful with the explanation of results, particularly regarding metabolic and pharmacokinetic interpretations, and always aim to compare the sum of species with the total arsenic concentration determined independently.