Blood analysis has been carried out routinely for decades. Normally, blood samples are taken from patients and sent to the laboratory for processing. Novel analytical methods are facilitated by fiberoptic sensors that are inserted into the human body or in contact with it. In the present work a spectral analysis of human blood serum was undertaken by fiberoptic evanescent wave spectroscopy (FEWS) using a Fourier transform infrared spectrometer (FTIR). A special cell based on IR transmitting non-toxic silver halide fibers was designed. Further improvements in the analysis were obtained by adopting some multivariate calibration techniques that have already been used in clinical chemistry. The partial least-square (PLS) algorithm was used to calculate the concentrations of urea, uric acid, calcium, cholesterol and total protein in human blood serum. Good agreement between our results and the ordinary chemical and enzymatic methods was reached. The average errors obtained were 28 mg/dl for cholesterol (about 16% of the average value), 9 mg/dl for urea (about 29% of the average value), 0.38 g/L for total protein (about 6% of the average value), 1.5 mg/dl for uric acid (about 29% of the average value) and 0.28 mg/dl for calcium (about 3% of the average value).