A method for the determination of the total iodine content in milk powder and human serum by inductively coupled plasma mass spectrometry is described. It is shown that the oxidation state of the analyte has a strong impact on the analytical performance. If iodine is present as iodide and nitric acid is used for acidification, the observed ion signal is not stable and samples spiked with iodide yield erroneous recoveries up to 750%. Hence, as the iodine in milk powder is present almost exclusively as iodide, destruction of the sample with nitric acid is not suitable. It was found that such an excessive spike recovery does not occur if ammonia solution instead of nitric acid is used as the diluent. The proposed sample preparation was tested by analysing three reference milk powders, namely Certified Reference Materials 150 and 151 Spiked Skim Milk Powder (from the Community Bureau of Reference) and Standard Reference Material 1549 Non-Fat Milk Powder (from the National Institute of Standards and Technology). Good agreement with the certified values was obtained, indicating that no significant errors are introduced. For human serum, samples treated with nitric acid or with ammonia solution gave results that were not significantly different from each other. Results are given for both a 'second-generation' biological reference material, Freeze-Dried Human Serum (University of Ghent), and for sera from healthy individuals.