Electron probe microanalysis has been a useful method for biological research for some years. Its advantages include the low background of x-ray spectra and range to include characteristic peaks of most biological elements. The traditional method of analyzing biological X-ray spectra filtering both the experimental and reference spectra with a top-hat filter, before performing a linear least squares fit, to remove the slowly varying background. We show that, as expected, this procedure introduces an additional error in the quantitation results, and reduces the sensitivity and precision of this method. Our experience with quantitating trace elements with EELS shows that direct fit of unfiltered spectra yields the highest sensitivity, when the data contain only statistical errors. Biological specimens usually contain high K and low Ca, and in X-ray spectra the K Kβ peak is overlapped with the Ca Kα peak. A small error in extracting the K signal will dramatically affect the quantitation of Ca. We demonstrate here the application of a direct fitting procedure, which focuses on the Ca and K region of the X-ray spectra (2800 ∼ 4900 eV), and yields smaller fitting uncertainties compared to the traditional filtered fit method in trace Ca quantitation.